A suspension material cross-cylinder transfer closed-loop control system and method

Abstract

The application discloses a kind of hanging material cross-cylinder transfer closed-loop control system and method, applied to pathological tissue section automatic dyeing processing equipment.The system includes hoisting execution mechanism, multi-source sensing monitoring component, fluid cleaning execution component and main control processor.The method includes: initial calibration to sensing monitoring component to establish mechanical background noise compensation atlas and acceleration-inertial impedance amplitude compensation coefficient characteristic mapping matrix;Dynamic tensile impedance and spatial posture dynamics electrical signal are collected when slide rack is hoisted and packaged as feature data frame;Combining atlas and matrix carries out noise elimination and motion inertia stripping, decouples and outputs absolute liquid wall mass;Multi-dimensional pulse purging parameter matrix is generated to drive fluid cleaning execution component to execute purging;During purging, the mass is monitored, and when it reaches the safety residual set threshold, the solenoid valve is turned off to block the airflow.The application eliminates reagent cross-contamination and achieves high-precision residual liquid closed-loop control.

Description

Technical Field

[0001] This application belongs to the field of control technology for automatic staining equipment for pathological tissue sections, specifically relating to a closed-loop control system and method for the trans-cylinder transfer of suspended materials. Background Technology

[0002] In the automated staining process of pathological tissue sections, slide holders need to carry a large number of slides and frequently transfer them between different reagent tanks (such as dewaxing tanks, staining tanks, dehydration tanks, and clearing tanks). Due to the tiny gaps between the slides, when the slide holder is lifted and removed from the current tank, a large amount of residual reagent solution is often trapped and encapsulated in the gaps due to capillary effect and liquid surface tension.

[0003] If these residual liquids are not effectively removed, they will continue to drip during subsequent cross-container transfer and entry processes, leading to serious cross-contamination of reagents. This will directly affect the purity of the chemical reaction in subsequent slide staining and may even cause the entire batch of pathological slides to be scrapped.

[0004] Traditional methods for removing residual liquid typically employ timed, pressure-controlled blind air blowing or simple mechanical vibration. However, the types of reagents used in pathological processing are numerous (such as xylene, ethanol of varying concentrations, staining solutions, etc.), and their physicochemical properties, such as density, dynamic viscosity, and surface tension, vary greatly. In actual industrial operation, the complex conditions caused by mechanical wear, ambient temperature rise, liquid level fluctuations, and continuous operation of multiple cylinders mean that static, blind air blowing control either results in insufficient purging (incomplete removal of residual liquid) or excessive purging (wasting airflow energy and even causing reagents to generate large amounts of aerosols that contaminate the equipment).

[0005] Meanwhile, existing control methods cannot accurately and decouple the calculation of the absolute mass of the actual liquid adhering to the slide holder in real time during equipment operation, resulting in a lack of closed-loop feedback benchmarks in the control system. Furthermore, the high-frequency, high-dimensional intelligent control algorithms running continuously on the underlying chip incur significant computational overhead, directly impacting the real-time response safety of the processing equipment. Therefore, achieving a closed-loop control system that can accurately identify the liquid adhering mass, adaptively adjust the purging sequence, and ensure a safe hard shutdown at the underlying level is a pressing technical bottleneck in this field. Summary of the Invention

[0006] To address the aforementioned technical issues, this application provides a closed-loop control system and method for the transfer of suspended materials across cylinders. This system can decouple and monitor the absolute liquid adhering to the wall with high precision, achieving intelligent adaptive precise pulse purging and closed-loop control, and completely preventing reagent cross-contamination.

[0007] In a first aspect, this application provides a closed-loop control system for the trans-cylinder transfer of suspended materials, applied in an automated staining and processing equipment for pathological tissue sections, comprising: The lifting actuator is used to drive the slide holder to perform cross-cylinder lifting and lowering operations in the vertical direction, and to switch the transfer position between different liquid cylinders in the horizontal direction. The lifting actuator includes a variable frequency lifting motor, a horizontal transfer transmission motor, a robotic arm, and a hook. A multi-source sensing and monitoring component is integrated at the physical connection between the hook and the slide holder, and is used to collect the dynamic tensile impedance electrical signal sequence of the hoisting actuator in the suspended section, as well as the spatial attitude dynamic electrical signal sequence of the robotic arm; A fluid scavenging actuator, located at the upper edge of each liquid cylinder, includes an external air source interface, a fluid collection chamber, and a nozzle assembly. An external high-pressure airflow enters the nozzle assembly through the external air source interface and the fluid collection chamber, forming an air curtain directed towards the slide gaps within the slide holder to remove adhering residual liquid from the slide gaps. The main control processor is electrically connected to the hoisting actuator, the multi-source sensing and monitoring component, and the fluid scavenging actuator, respectively. The main control processor includes: The signal amplitude decoupling unit is used to perform time-series alignment, time-frequency joint conversion, parasitic mechanical noise filtering, and physical stripping of inertial additional impedance on the dynamic stretching impedance electrical signal sequence and the spatial attitude dynamic electrical signal sequence, so as to output the absolute liquid wall mass of the liquid encased on the glass slide holder. An unsupervised adaptive control unit for operating condition characteristics is used to retrieve reagent physical prior constants based on the current physicochemical characteristics of the liquid leaving the cylinder, and to achieve spatial alignment and temporal expansion of operating condition characteristics by minimizing the joint loss function of the physical field, so as to calculate and output a multidimensional pulse purging parameter matrix; and The event-triggered prediction execution unit is used to construct a time-series state space vector in real time based on the absolute liquid wall mass, allocate optimization computing power through dual-channel event-triggered logic, and perform upper-layer forced safety limit coverage monitoring. When the safety residue set threshold is reached, the fluid scavenging execution component is shut down.

[0008] As a preferred embodiment, the multi-source sensing and monitoring component includes: a resistive strain sensor integrated at the physical connection between the hook and the slide holder, used to acquire and output the dynamic tensile impedance electrical signal sequence, and the resistive strain sensor adopts a bipolar differential compensation design, using compensation electrodes connected in parallel at both ends of the bridge to cancel temperature rise common-mode interference; and a triaxial inertial spatial acceleration and angular velocity sensor embedded in the robotic arm, used to synchronously acquire the instantaneous vibration acceleration components and angular velocity yaw components of the robotic arm in three spatial axes, so as to generate and output the spatial attitude dynamic electrical signal sequence.

[0009] As a preferred embodiment, the fluid collection chamber is specifically configured as follows: an annular closed cavity surrounding the upper edge of the liquid cylinder; the nozzle assembly consists of a ring of micro-pulse spray holes formed on the inner wall of the annular closed cavity, used to spray the high-pressure airflow after stabilization in the annular closed cavity centripetally along the preset airflow spray angle to form the air curtain; a solenoid valve and a proportional pressure regulating valve are connected in series on the pipeline of the external air source interface, and the control terminals of the solenoid valve and the proportional pressure regulating valve are electrically connected to the main control processor respectively.

[0010] As a preferred embodiment, the variable frequency lifting motor controls the slide holder to perform a vertically upward cylinder detachment action and a vertically downward cylinder immersion action by switching between forward and reverse operation states; the horizontal transfer transmission motor drives the robotic arm to perform translational movement along a preset horizontal track to achieve precise axial alignment of the slide holder between different liquid cylinders.

[0011] Secondly, this application provides a closed-loop control method for the transfer of suspended materials across cylinders, applied in the aforementioned control system, comprising the following steps: The multi-source sensing and monitoring components were initially calibrated, and a mechanical background noise compensation spectrum and an acceleration-inertial impedance amplitude compensation coefficient feature mapping matrix were established. The hoisting actuator is driven to lift the slide holder from the current liquid cylinder upwards, and the dynamic tensile impedance electrical signal sequence and the spatial attitude dynamic electrical signal sequence during the hoisting process are collected and encapsulated into a multi-dimensional time-series feature data frame. Combining the mechanical background noise compensation spectrum and the acceleration-inertial impedance amplitude compensation coefficient feature mapping matrix, the multidimensional time-series feature data frame is subjected to time-frequency noise reduction and motion inertia stripping, and the absolute liquid wall mass of the liquid encased on the slide holder is decoupled and output. Based on the absolute liquid wall mass, an unsupervised working condition feature migration mechanism is used to adaptively solve the problem by combining the current physicochemical and working condition data of the liquid cylinder, generating a multi-dimensional pulse purging parameter matrix, and driving the fluid purging execution component to perform the purging action; wherein, the multi-dimensional pulse purging parameter matrix is ​​used to monitor the absolute liquid wall mass during the purging process, and when the absolute liquid wall mass reaches or falls below the safe residual set threshold, the solenoid valve configured in the fluid purging execution component is controlled to shut off to block the airflow.

[0012] The beneficial effects of this application are as follows: Through initial calibration and motion inertia stripping, this application achieves precise decoupling and online identification of the absolute liquid adhesion mass on the surface and in the gaps of the slide holder; furthermore, by introducing an unsupervised operating condition feature transfer mechanism at the upper layer, and combining the current physicochemical and operating condition data of the liquid cylinder for adaptive solution, the cost of tedious manual modeling for each reagent is eliminated, dynamically generating a multi-dimensional pulse purging parameter matrix that fits the actual physical field; simultaneously, this matrix is ​​used at the middle and lower layers to monitor the adhesion mass during purging, and controls the solenoid valve to perform a safety shutdown when it falls below the safe residual threshold. This application significantly reduces the computing power and power consumption of the control chip through signal multiplexing when the system is stable, and fundamentally eliminates reagent cross-contamination, ensuring safe, low-power, and high-precision closed-loop control of the fully automated pathological staining process. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of a closed-loop control system for the transfer of suspended materials across cylinders, provided in an embodiment of this application. Figure 2 This is a schematic flowchart of a closed-loop control method for the transfer of suspended materials across cylinders, provided in an embodiment of this application. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] This application provides a closed-loop control system for the trans-cylinder transfer of suspended materials, applied in an automated staining and processing equipment for pathological tissue sections. (See also...) Figure 1 , Figure 1This is a schematic diagram of a closed-loop control system for suspended material transfer across cylinders, provided in an embodiment of this application. The closed-loop control system mainly includes: a hoisting actuator 10, a multi-source sensing monitoring component 20, a fluid scavenging actuator 30, and a main control processor (not shown in the diagram) electrically connected to them. To make the technical solution and beneficial effects of this application more intuitive and clear, the specific mechanisms and cooperation of each component are explained below, combining the physical configuration of the core system mechanism and the signal decoupling control logic: The lifting actuator 10, serving as the system's power and load-bearing main body, drives the slide holder to perform high-precision cross-cylinder transfer lifting and lowering operations in the vertical direction. After detaching from the cylinder, it drives the slide holder to precisely switch between different liquid cylinders in the horizontal direction. In this embodiment, the main hardware architecture of the lifting actuator 10 includes: a variable frequency lifting motor, a horizontal transfer transmission motor, a robotic arm, and a hook. Wherein: A variable frequency lifting motor, by switching between forward and reverse operation, controls the slide holder to perform a vertically upward cylinder release action and a vertically downward cylinder immersion action. The variable frequency lifting motor is a variable frequency brake motor with a high transmission ratio and self-locking function. Its output shaft is physically connected to the robotic arm via a synchronous pulley, lead screw drive pair, or gear and rack mechanism. By switching the forward and reverse operation states of the variable frequency lifting motor, its output torque and speed are adjusted in real time, thereby precisely controlling the slide holder below to sequentially perform a vertically upward variable speed cylinder release action or a vertically downward cylinder immersion action.

[0018] The horizontal transfer drive motor is installed on the side of the horizontal track of the crossbeam at the top of the equipment. It drives the entire robotic arm to move along the preset horizontal transverse track through the gantry slider mechanism. Its positioning accuracy is controlled within ±0.1mm, thereby achieving precise axial alignment of the slide holder between different functional liquid cylinders.

[0019] The robotic arm is constructed from high-rigidity aerospace-grade aluminum alloy or corrosion-resistant stainless steel hollow tubing, extending vertically downwards. A hook for carrying materials is physically fixed at its end. This hook is physically suspended from a slide holder containing pathological slides via a flexible suspension component. Due to this flexible connection, the slide holder will inevitably experience spatial sway and mechanical flutter during transient processes such as vertical lifting or horizontal acceleration.

[0020] To accurately and in real-time extract the mass of trace residual liquid carried on the surface of the slide holder under the complex conditions of flexible suspension and dynamic vibration, this application seamlessly integrates a multi-source sensing monitoring component 20 onto the lifting actuator 10. The specific sensor spatial arrangement structure of the multi-source sensing monitoring component 20 is as follows: A resistive strain gauge sensor, integrated at the physical connection between the hook and the slide holder, is used to acquire and output the dynamic tensile impedance signal sequence. In this embodiment, the resistive strain gauge sensor is a miniature high-precision resistive strain gauge patch, directly physically bonded or welded to the neck of the physical force-bearing connection between the hook and the slide holder. When the slide holder is vertically suspended, the minute tensile elastic deformation generated at the hook neck is directly transmitted to the strain gauge, causing a change in its resistance value, thereby outputting a continuous dynamic tensile impedance signal sequence.

[0021] Given the significant localized temperature rise within pathological staining equipment, temperature fluctuations can cause thermal output errors in strain gauges. To address this, the resistive strain sensor employs a bipolar differential compensation design. A working strain gauge is positioned along the principal force direction of the hook, while symmetrical temperature compensation electrodes are connected in parallel in a region adjacent to this position but not subjected to tension. When the ambient temperature drifts, the working electrode and the compensation electrode generate resistance changes of the same direction and amplitude. These changes are completely offset by the compensation electrodes connected in parallel across the bridge, ensuring the purity of the impedance spectrum from the outset.

[0022] A three-axis inertial spatial acceleration and angular velocity sensor is embedded in the robotic arm to synchronously acquire the instantaneous vibration acceleration components and angular velocity yaw components of the robotic arm in three spatial axes, thereby generating and outputting the spatial attitude dynamics electrical signal sequence. In this embodiment, a six-axis micromechanical inertial measurement unit chip with a high sampling rate is selected, and it is firmly embedded and closely attached to the middle surface of the vertical cantilever of the robotic arm in a rigid package. This sensor can accurately capture the low-frequency oscillation waveforms caused by the commutation of the variable frequency lifting motor and the braking of the horizontal transfer drive motor.

[0023] The fluid scavenging actuator 30 is horizontally and fixedly embedded at the uppermost open edge of each liquid cylinder, forming a dynamic gas interception barrier. Based on the narrow-slit fluid acceleration principle of the air knife, the specific mechanical and pneumatic configuration of the fluid scavenging actuator 30 is as follows: The fluid collection cavity is specifically configured as follows: an annular closed cavity surrounding the upper edge of the liquid cylinder. This cavity is precision-machined from an acid and alkali resistant composite polymer material or anodized aluminum. Its structure perfectly matches the upper opening contour of the liquid cylinder, forming a seamless closed annular double-layered hollow cavity. An external air source interface is provided on the outer wall of this cavity. After the external high-pressure airflow enters this cavity, it undergoes circumferential flow splitting and pressure equalization within the closed annular cavity, forming a stable high-pressure stabilizing airflow field.

[0024] The nozzle assembly consists of a ring of micro-pulse nozzles formed on the inner wall of the annular closed cavity. These nozzles, acting as the nozzle assembly, are uniformly and continuously formed on the inner wall along the circumference, with a predetermined downward tilt angle between their spray centerline and the vertical centerline. When the high-pressure, stabilized airflow is ejected centripetally through these micro-pulse nozzles, according to the principles of fluid amplification and laminar flow aggregation, it interweaves and merges into a continuous, dense, and transversely tangential centripetal interwoven air curtain at the center of the cavity. When a slide holder carrying densely packed pathological tissue sections passes upward through this curtain, the high-speed shearing force of the air curtain, acting like a "scalpel," directly penetrates and removes residual liquid adhering to the narrow gap between adjacent slides.

[0025] A solenoid valve and a proportional pressure regulating valve are connected in series on the pipeline of the external air source interface. The control terminals of the solenoid valve and the proportional pressure regulating valve are electrically connected to the main control processor. The solenoid valve has millisecond-level response capability and is used to control the opening and closing of the air curtain and the high-frequency pulse frequency; the proportional pressure regulating valve is used to finely and linearly adjust the pressure of the injected airflow.

[0026] The main control processor is electrically connected to the hoisting actuator 10, the multi-source sensing and monitoring component 20, and the fluid sweeping execution component 30, respectively. The main control processor integrates a memory, a signal amplitude decoupling unit, an unsupervised operating condition characteristic adaptive control unit, and an event-triggered predictive execution unit. During operation, it executes the following closed-loop process: The signal amplitude decoupling unit is used to perform time-series alignment, time-frequency joint conversion, parasitic mechanical noise filtering, and physical stripping of inertial additional impedance on the dynamic tensile impedance electrical signal sequence and the spatial attitude dynamic electrical signal sequence, so as to output the absolute liquid wall mass of the liquid encased on the slide holder. Since the inertial force generated by the robotic arm movement, the noise caused by the slide rail friction, and the centripetal force caused by the slide holder swing are all superimposed on the hook's tensile impedance signal, this unit first uses a cross-correlation function to perform reference time-series alignment on the two sets of signals, and then performs time-frequency joint conversion through Fast Fourier Transform or Wavelet Transform. Parasitic mechanical noise is filtered out by using a pre-retrieved system inherent mechanical vibration frequency spectrum. Finally, a dynamic Newton-Euler dynamic equation is constructed using the spatial attitude angular velocity and acceleration to physically strip the inertial additional impedance from the spectrum. The stripped net impedance value is directly mapped to a pure gravity load, thereby accurately calculating and outputting the absolute liquid wall mass.

[0027] The unsupervised adaptive control unit for operating conditions retrieves prior physical constants of the reagents based on the physicochemical characteristics of the liquid leaving the cylinder. It then achieves spatial alignment and temporal expansion of the operating condition characteristics by minimizing the joint loss function of the physical fields, thereby calculating and outputting a multidimensional pulse purging parameter matrix. Due to the significant differences in the physicochemical properties of reagents in different cylinders, this unit automatically retrieves the corresponding prior physical constants of the reagents from its memory based on the current operating condition of the robotic arm. By running an unsupervised adaptive algorithm, it minimizes the joint loss function of the physical fields, which includes fluid dynamics boundary conditions, achieving spatial alignment and temporal expansion of the current operating condition characteristics. This allows it to calculate the most suitable multidimensional pulse purging parameter matrix for the current liquid stripping process. This matrix explicitly defines the pulse switching frequency and duty cycle of the solenoid valve, as well as the set target pressure value of the proportional pressure regulating valve. The main control processor then drives the fluid purging execution component 30 to generate a precisely proportioned pulse air curtain according to these matrix parameters.

[0028] The event-triggered prediction execution unit is used to construct a time-series state space vector in real time based on the absolute liquid wall mass. It allocates optimization computing power through dual-channel event-triggered logic and executes upper-level forced safety limit coverage monitoring. When the set safety residual threshold is reached, it shuts down the fluid purging execution component 30. During the purging process, the absolute liquid wall mass decreases rapidly over time. This unit combines the current mass, robotic arm lifting speed, and air curtain pressure in real time to construct a time-series state space vector. The unit uses dual-channel event-triggered logic. When the mass decrease gradient is gradual, the calculation frequency is reduced to save computing power. When the mass is detected to be close to the critical point, a high-frequency prediction channel is triggered instantaneously. Simultaneously, the unit executes upper-level forced safety limit coverage monitoring. Once the absolute liquid wall mass is predicted or detected to have decreased to the preset safety residual threshold, the unit immediately sends a zero-voltage shutdown command to the solenoid valve, completely shutting down the fluid purging execution component 30 and stopping the air jet.

[0029] Through the mechanical cooperation between the above components, the input of multi-source sensor signals, the real-time decoupling of the main control algorithm and the adaptive adjustment of parameters, this application has successfully constructed a fully closed-loop intelligent control system covering "material lifting - dynamic sensing - laminar flow purging - precise adaptive shutdown".

[0030] Based on the foregoing detailed description of the hardware architecture, physical connections, and sensor deployment structure of the suspended material transfer closed-loop control system, it is clear that the hardware components do not exist in isolation during actual operation, but are organically integrated through a highly coordinated digital signal closed-loop control logic. To further demonstrate how this invention achieves precise removal of adhering residual liquid during the material transfer process based on the aforementioned hardware foundation, this invention provides a suspended material transfer closed-loop control method.

[0031] See Figure 2 , Figure 2 This is a schematic flowchart of a closed-loop control method for the transfer of suspended materials across cylinders, provided in an embodiment of this application. It includes the following steps: Step S1: Perform initial calibration on the multi-source sensing and monitoring components, and establish a mechanical background noise compensation spectrum and an acceleration-inertial impedance amplitude compensation coefficient feature mapping matrix.

[0032] During the initial power-on of the equipment or the static interval between batch transitions, the main control processor invokes its internal calibration and firmware program. To achieve high-precision zero-point elimination and removal of mechanical spurious variables, the main control processor continuously acquires the initial impedance value of the empty slide holder at a reference sampling rate of 500Hz to 2000Hz (preferably 1000Hz) using its integrated 24-bit high-precision analog-to-digital converter (ADC). The system determines the static reference gravity impedance value (typically between 10Ω and 500Ω) by continuously acquiring 100 to 500 sampling points and calculating their arithmetic mean. Then, by executing an arithmetic mean filtering algorithm to eliminate white noise interference, the static reference gravity impedance value (in ohms) is calculated and written as a static zero-point bias parameter into a non-volatile specific sector of the EEPROM memory, serving as the reference subtraction for all subsequent gravity calculations.

[0033] Subsequently, the main control processor sends stepped pulse commands to the driver of the variable frequency lifting motor, controlling the variable frequency lifting motor to drive the empty slide holder vertically upward at preset multi-level uniform speed gradients (such as 0.05m / s, 0.1m / s, and 0.3m / s stepped speeds). Within each uniform speed running segment, the resistive strain gauge collects stray electrical signals caused by friction from the surface micro-roughness of the stainless steel vertical guide rail, the impact of the synchronous belt micro-step meshing, and the resonance of the machine body and casing. The main control processor performs high-pass filtering on these stray electrical signals, extracts the energy density distribution across the entire frequency band, and superimposes and maps them in the time and frequency domains, thereby constructing a mechanical background noise compensation spectrum in memory with frequency (Hz) and amplitude voltage (V) as dual-axis coordinates.

[0034] Then, the main control processor drives the variable frequency lifting motor to control the slide holder to perform a uniformly accelerated vertical upward motion (acceleration set to 0.5 m / s²). 2 up to 1.5m / s 2During the rapid ascent under hypergravity, the main control processor utilizes a six-axis IMU (Inertial Measurement Unit) chip integrated on the external SPI bus to simultaneously read real-time vertical acceleration values ​​(typically between 1.05g and 1.15g) and the tensile impedance fluctuations caused by the hypergravity from the resistive strain gauge. Since the impedance change and acceleration physically satisfy the strain gauge deformation tensor relationship, the main control processor internally runs a least-squares fitting program to fit the first and second derivatives of acceleration with the impedance increment using a partial differential matrix, thereby solving and fitting the characteristic mapping matrix of the acceleration-inertial impedance amplitude compensation coefficient. This matrix is ​​stored in memory as a double-precision floating-point coefficient matrix, allowing any dynamic acceleration input to be directly converted into its corresponding "pseudo-gravity impedance increment" by multiplying it.

[0035] To combat sensor thermal drift caused by temperature fluctuations within the staining chamber over time, the main control processor utilizes the brief process interval (typically 3 to 5 seconds) during which the slide holder is immersed in the liquid tank after each process is completed to trigger ADC sampling again. The impedance value at this time is compared with the historical static bias, and the static zero-point bias parameters are dynamically updated to achieve dynamic reference self-calibration of the entire machine.

[0036] Step S2: Drive the hoisting actuator to lift the slide holder from the current liquid cylinder upwards, and collect the dynamic tensile impedance electrical signal sequence and the spatial attitude dynamic electrical signal sequence during the hoisting process, and encapsulate them into a multi-dimensional time-series feature data frame.

[0037] In actual operation, the main control processor drives the main control inverter bridge of the variable frequency drive motor through PWM (Pulse Width Modulation) signals. During the initial acceleration lifting phase, the motor's output torque increases linearly, driving the glass slide holder at a first preset acceleration (e.g., 0.6 m / s²). 2 Up to 1.2 m / s 2 The slide holder accelerates vertically upwards and detaches from the liquid surface. When the lower edge of the slide holder is determined by the liquid level sensor to have completely detached from the liquid surface, the inverter switches to constant frequency control and enters the uniform speed operation phase, driving the slide holder to maintain a uniform vertical upward movement at a first preset speed (e.g., 0.2m / s to 0.4m / s).

[0038] At the moment when the variable frequency lifting motor receives the start control command, the main control processor uses this as a unified time zero point and uses an internal hardware counter to synchronously acquire the dynamic tensile impedance electrical signal sequence and the spatial attitude dynamics electrical signal sequence (i.e., triaxial acceleration and angular velocity signals) output by the multi-source sensing and monitoring component 20 at a high preset sampling rate of 1kHz to 5kHz (preferably 2kHz).

[0039] To ensure strict correspondence between signals from different channels on the time axis, the main control processor performs timing alignment on the two synchronously acquired electrical signal sequences based on the synchronization time reference corresponding to the preset sampling rate (with accuracy controlled within 1 microsecond). Subsequently, the main control processor adopts a sliding time window algorithm, setting a fixed time window length of 128ms to 512ms (corresponding to 256 or 512 sampling points), and slides forward 10% to 25% of the window each time, extracting the aligned electrical signals within the time window and encapsulating them into a multi-dimensional timing feature data frame with sequential timing logic.

[0040] During this encapsulation process, the main control processor synchronously retrieves the physical prior constants of the reagent currently leaving the liquid cylinder (e.g., the standard density of eosin reagent at the current temperature of 25°C is 0.8 g / cm³, and the dynamic viscosity is 1.2 mPa·s), and directly injects these physical prior constants as key feature weight coefficients into the specific message channel of the multidimensional time-series feature data frame, so that the data frame itself carries the physical properties of the current fluid condition. Step S3: Combining the mechanical background noise compensation spectrum and the acceleration-inertial impedance amplitude compensation coefficient feature mapping matrix, perform time-frequency noise reduction and motion inertia stripping on the multi-dimensional time-series feature data frame, and decouple and output the absolute liquid wall mass of the liquid encased on the slide holder.

[0041] After receiving the encapsulated multi-dimensional timing feature data frame, the main control processor calls its internal digital signal processing core and, in conjunction with the parameters generated during the calibration phase, performs a fully digital decoupling operation: First, the system performs a discrete Fourier transform on the dynamic stretching impedance electrical signal sequence in the multidimensional time-series feature data frame, and uses the time window overlap shifting method (time window overlap rate set to 50% to 75%) to slide on the time axis for spectrum analysis, transforming it into an absolute quality feature spectrum with both time and frequency domain dimensions.

[0042] Next, the main control processor reads the pre-stored mechanical background noise compensation spectrum and performs a spectral comparison within the absolute quality characteristic spectrum. It locates the mechanical resonance noise value within the first preset frequency band (set between 60Hz and 150Hz, corresponding to the high-frequency band of the robotic arm guide rail and motor micro-stepping vibration), and performs adaptive band-stop suppression (attenuation exceeding 30dB) on the resonance value within this frequency band, thereby filtering out the mechanical resonance interference signal. In the filtered spectrum, the system extracts and retains the absolute liquid spectral energy component below the second preset frequency band threshold (set between 0Hz and 12Hz, a quasi-static extremely low frequency band), completely eliminating high-frequency mechanical noise.

[0043] Then, the main control processor extracts the real-time vertical acceleration vector from the spatial attitude dynamics electrical signal during the same period, and multiplies it by the pre-calibrated acceleration-inertial impedance amplitude compensation coefficient characteristic mapping matrix to calculate the inertial additional impedance stretching component (the signal amplitude is usually a pseudo-variable of 0.5V to 2.0V) generated by the acceleration motion of the robotic arm. The main control processor controls the retained absolute liquid spectral energy component to subtract this inertial additional impedance stretching component in real time, thereby physically completely separating the gravitational stretching response and obtaining the inertial compensated signal to be calculated.

[0044] Finally, the inertial-compensated signal to be processed is input into a multi-path parallel characteristic bottleneck impedance calculation network constructed inside the main control processor. This network performs parallel decoupling simultaneously through four parallel characteristic processing physical branch channels: The first feature processing physical branch channel: specifically extracts the slow gravitational field jump features caused by the net mass of the reagent (usually decreasing between 1g and 50g); The second feature processing physical branch channel: the transient non-gravitational oscillation component (frequency distribution between 1Hz and 5Hz) generated by the surface undulation during the upward shaking of the special decoupling reagent. The third feature processing physical branch channel: specifically filters out the robotic arm sliding friction force that manifests as sudden spikes due to the axial micro-sliding of the robotic arm slide; The fourth feature processing physical branch channel: specifically analyzes the tangential impact reaction force generated by the nozzle air sweep pressure on the glass slide (this reaction force is converted into a pseudo-gravity signal in the vertical direction).

[0045] The feature data output from the four independent channels are merged in dimensions through a splicing layer and normalized by a weighted average fusion algorithm. Finally, the decoupled output is the true and pure absolute liquid wall mass that eliminates motion interference, friction interference and wind pressure reaction force interference. Its mass calculation accuracy can be stably maintained between ±5mg and ±10mg.

[0046] Step S4: Based on the absolute liquid wall mass, an adaptive solution is performed using an unsupervised operating condition feature migration mechanism combined with the current physicochemical and operating condition data of the liquid cylinder to generate a multi-dimensional pulse purging parameter matrix, and drive the fluid purging execution component to perform the purging action; wherein, the multi-dimensional pulse purging parameter matrix is ​​used to monitor the absolute liquid wall mass during the purging process, and when the absolute liquid wall mass reaches or falls below the safe residual set threshold, the solenoid valve configured in the fluid purging execution component is controlled to shut off to block the airflow.

[0047] After decoupling the absolute liquid wall mass, the system initiates dynamic purging decision-making. The main control processor reads the calibration condition characteristic data (based on a nominal environment of 20℃) pre-stored in the memory and obtains the target reagent type identifier of the current working cylinder. The corresponding basic physical property parameter vector, consisting of reagent density, dynamic viscosity, and surface tension coefficient, is used as boundary constraint terms and input into the unsupervised physical and chemical condition characteristic migration adaptive control network.

[0048] Utilizing the unsupervised adaptive control network for physicochemical condition feature transfer, the system transfers the calibration conditions from the standard environment to the actual operating conditions of the current liquid cylinder. The network compensates for the time-varying shifts in the environment and reagents by minimizing the joint loss function of the physical field, which includes control bias and the time-varying physicochemical characteristics of the reagents. This achieves feature space reconstruction alignment between the calibration and actual operating conditions, and outputs the control decision parameter vector for the current purging phase (containing the target total pressure value of 0.2 MPa to 0.6 MPa, the target pulse frequency of 10 Hz to 100 Hz, and the target duty cycle of 30% to 70%) and the corresponding confidence score (ranging from 0.00 to 1.00).

[0049] To improve decision-making stability, the system introduces a temporal adaptive iterative mechanism. Utilizing the temporal consistency constraint in the joint loss function of the physical fields, the system dynamically weights and corrects the credibility score. When the corrected credibility score meets a preset credibility threshold (set to 0.80 to 0.90, preferably 0.85), the system determines that the current network migration result is highly accurate. It then uses the actual motion trajectory features corresponding to the control decision parameter vector as update features, injecting them in reverse and overwriting the calibration condition feature data. This achieves dynamic temporal expansion of the calibration feature space along the time axis and online incremental learning.

[0050] Based on an unsupervised adaptive control network that has completed feature space reconstruction and alignment, the system combines the real-time input of the absolute liquid wall mass to map and solve the control decision parameter vector, ultimately generating a multi-dimensional pulse purging parameter matrix that conforms to the current physical characteristics of the liquid cylinder. The main control processor converts this matrix into electrical signal commands to drive the proportional pressure regulating valve (outputting a standard control current of 4mA to 20mA) and the solenoid valve (outputting a 24V pulse voltage logic signal) in the fluid purging execution component 30, finely adjusting the total airflow pressure and executing microsecond-level (on / off response delay ≤2ms) timing on / off actions to eject a dynamic centripetal air curtain.

[0051] During the purging action performed by the fluid purging execution component 30, in order to respond to the residual liquid removal status within milliseconds and protect the sliced ​​tissue, the main control processor calls the internal event-triggered predictive execution unit.

[0052] The main control processor, based on the absolute liquid adhesion mass output in real time from the DSP core, continuously calculates the mass deviation between the current mass and the safety setpoint, as well as the rate of change of the current mass (in mg / ms). These two indicators together constitute a real-time state feature vector reflecting the dynamic sliding trend of the liquid film. The event-triggered prediction execution unit, based on the aforementioned real-time state features, performs hierarchical control logic distribution on this vector within each short sampling period (set to 1ms to 5ms, preferably 2ms): When the real-time status characteristics meet the first channel trigger condition (i.e., the mass change rate is within the preset stable range of -0.5mg / ms to 0mg / ms, and the deviation converges linearly, indicating that the purging condition fully conforms to the predetermined slope), in order to save the chip core computing power, the system no longer calls the network again to perform complex calculations, but triggers the first channel, sequentially retrieves and directly reuses the control signals in the multi-dimensional pulse purging parameter matrix of the previous cycle from the SRAM storage buffer configured in the event trigger prediction execution unit, and maintains the predetermined pulse output of the solenoid valve.

[0053] When the real-time state characteristics meet the second channel trigger condition (i.e., a sudden change in liquid film peeling, a mass change rate exceeding the above-mentioned stable range, or a stagnation in the decrease of mass deviation, indicating that there may be local liquid bridge adhesion on the slide), the system instantly activates time-domain predictive control calculation, performs rolling optimization calculation in the long time domain of the next 3 to 8 prediction steps, dynamically adjusts the timing of the multi-dimensional pulse purging parameter matrix online, updates the storage buffer and immediately outputs the updated control sequence, and dynamically adjusts the pulse duty cycle or target pressure of the airflow.

[0054] Throughout the purging process, the system implements high-priority online real-time monitoring through the dual-channel upper-level safety limit coverage monitor in the event-triggered predictive execution unit. Once it is determined that the absolute liquid residue mass has reached or fallen below the preset safety residual threshold (set between 5mg and 15mg, preferably 10mg, indicating that the attached residual liquid has been completely removed, and continued purging will cause the sample to lose water, dry out, and crack), or if it is detected that the control signal in the storage buffer has reached the preset time-domain limit protection value (set between 100 and 300 steps, preferably 150 steps, triggering foolproof and timeout protection) due to the smooth multiplexing of the first channel, the safety limit coverage monitor will forcibly override all underlying control logic and forcibly output a 0V zero-bias voltage signal to the solenoid valve in the fluid purging execution component 30. The solenoid valve will instantly and completely shut off within the preset hardware shutdown delay (≤1ms shutdown delay), completely blocking the high-pressure airflow and preventing the slide sample from being excessively dried.

[0055] When the solenoid valve is completely shut off and the high-pressure air curtain disappears completely, it marks the successful completion of the dynamic liquid removal process at the current liquid cylinder station.

[0056] The main control processor then sends synchronous translation and lifting combined action commands to the multi-axis motion control bus (such as EtherCAT bus). The control system drives the horizontal transfer drive motor and the variable frequency lifting motor in the lifting actuator 10 to operate in strict accordance with the preset spatial composite trajectory curve: the variable frequency lifting motor continues to lift the robotic arm vertically upward to the safe suspension height, while the horizontal transfer drive motor drives the entire gantry slider mechanism, carrying a clean glass slide holder with no risk of liquid cross-drip contamination, to move horizontally from the current workstation and smoothly and without vibration to the preset target workstation. The main control processor compares the preset process formula flow and accurately sends the glass slide holder into the next target liquid cylinder for the next dye immersion, or directly into the drying chamber for high-temperature curing and drying, or to the temporary storage area and unloading position, thereby completing the closed-loop intelligent control of the cross-cylinder transfer of suspended materials between multiple processing workstations.

[0057] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0058] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following drawings denote similar items; therefore, once an item is defined in one drawing, it need not be further discussed in subsequent drawings.

[0059] In the description of this application, it should be understood that the orientation or positional relationship indicated by directional terms such as "front, back, up, down, left, right", "horizontal, vertical, horizontal" and "top, bottom" is usually based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing this application and simplifying the description. Unless otherwise stated, these directional terms do not indicate or imply that the device or element referred to must have a specific orientation or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the scope of protection of this application; the directional terms "inner" and "outer" refer to the inner and outer contours relative to the outline of each component itself.

[0060] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.

[0061] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore cannot be construed as limiting the scope of protection of this application.

[0062] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A closed-loop control system for the trans-cylinder transfer of suspended materials, applied in an automated staining and processing equipment for pathological tissue sections, characterized in that, include: The lifting actuator is used to drive the slide holder to perform cross-cylinder lifting and lowering operations in the vertical direction, and to switch the transfer position between different liquid cylinders in the horizontal direction. The lifting actuator includes a variable frequency lifting motor, a horizontal transfer transmission motor, a robotic arm, and a hook. A multi-source sensing and monitoring component is configured on the hoisting actuator to collect the dynamic tensile impedance electrical signal sequence of the hoisting actuator in the suspended section, as well as the spatial attitude dynamic electrical signal sequence of the robotic arm. A fluid scavenging actuator, located at the upper edge of each liquid cylinder, includes an external air source interface, a fluid collection chamber, and a nozzle assembly. An external high-pressure airflow enters the nozzle assembly through the external air source interface and the fluid collection chamber, forming an air curtain directed towards the slide gaps within the slide holder to remove adhering residual liquid from the slide gaps. The main control processor is electrically connected to the hoisting actuator, the multi-source sensing and monitoring component, and the fluid scavenging actuator, respectively. The main control processor includes: Memory; The signal amplitude decoupling unit is used to perform time-series alignment, time-frequency joint conversion, parasitic mechanical noise filtering, and physical stripping of inertial additional impedance on the dynamic stretching impedance electrical signal sequence and the spatial attitude dynamic electrical signal sequence, so as to output the absolute liquid wall mass of the liquid encased on the glass slide holder. An unsupervised adaptive control unit for operating condition characteristics is used to retrieve reagent physical prior constants based on the current physicochemical characteristics of the liquid leaving the cylinder, and to achieve spatial alignment and temporal expansion of operating condition characteristics by minimizing the joint loss function of the physical field, so as to calculate and output a multidimensional pulse purging parameter matrix; and The event-triggered prediction execution unit is used to construct a time-series state space vector in real time based on the absolute liquid wall mass, allocate optimization computing power through dual-channel event-triggered logic, and perform upper-layer forced safety limit coverage monitoring. When the safety residue set threshold is reached, the fluid scavenging execution component is shut down. A resistive strain gauge, integrated at the physical connection between the hook and the slide holder, is used to acquire and output the dynamic tensile impedance electrical signal sequence. The resistive strain gauge employs a bipolar differential compensation design, using compensation electrodes connected in parallel across the bridge to cancel common-mode interference caused by temperature rise. A three-axis inertial space acceleration and angular velocity sensor is embedded in the robotic arm to synchronously acquire the instantaneous vibration acceleration components and angular velocity yaw components of the robotic arm in three spatial axes, so as to generate and output the spatial attitude dynamics electrical signal sequence.

2. The control system according to claim 1, characterized in that, The fluid collection cavity is specifically configured as follows: an annular closed cavity surrounding the upper edge of the liquid cylinder; The nozzle assembly consists of a ring of micro-pulse spray holes formed on the inner sidewall of the annular closed cavity, which is used to spray the high-pressure airflow after stabilization in the annular closed cavity outward along the preset airflow spray angle to form the air curtain. A solenoid valve and a proportional pressure regulating valve are connected in series on the pipeline of the external gas source interface, and the control terminals of the solenoid valve and the proportional pressure regulating valve are electrically connected to the main control processor.

3. The control system according to claim 1, characterized in that, The variable frequency lifting motor controls the slide holder to perform a vertically upward detachment from the cylinder and a vertically downward immersion into the cylinder by switching between forward and reverse operation states. The horizontal transfer drive motor drives the robotic arm to perform translational movements along a preset horizontal track, thereby achieving precise axial alignment of the slide holder between different liquid cylinders.

4. A closed-loop control method for the transfer of suspended materials across cylinders, characterized in that, The method, when applied to a control system as described in any one of claims 1 to 4, includes the following steps: The multi-source sensing and monitoring components were initially calibrated, and a mechanical background noise compensation spectrum and an acceleration-inertial impedance amplitude compensation coefficient feature mapping matrix were established. The hoisting actuator is driven to lift the slide holder from the current liquid cylinder upwards, and the dynamic tensile impedance electrical signal sequence and the spatial attitude dynamic electrical signal sequence during the hoisting process are collected and encapsulated into a multi-dimensional time-series feature data frame. Combining the mechanical background noise compensation spectrum and the acceleration-inertial impedance amplitude compensation coefficient feature mapping matrix, the multidimensional time-series feature data frame is subjected to time-frequency noise reduction and motion inertia stripping, and the absolute liquid wall mass of the liquid encased on the slide holder is decoupled and output. Based on the absolute liquid wall mass, an unsupervised working condition feature migration mechanism is used to adaptively solve the problem by combining the current physicochemical and working condition data of the liquid cylinder, generating a multi-dimensional pulse purging parameter matrix, and driving the fluid purging execution component to perform the purging action; wherein, the multi-dimensional pulse purging parameter matrix is ​​used to monitor the absolute liquid wall mass during the purging process, and when the absolute liquid wall mass reaches or falls below the safe residual set threshold, the solenoid valve configured in the fluid purging execution component is controlled to shut off to block the airflow.

5. The control method according to claim 4, characterized in that, The steps of initially calibrating the multi-source sensing and monitoring components and establishing the mechanical background noise compensation spectrum and the characteristic mapping matrix of the acceleration-inertial impedance amplitude compensation coefficient include: When the hoisting actuator is stationary, the initial impedance value of the unloaded slide holder is continuously collected, and the static reference gravity impedance value is calculated and determined as a static zero-point bias parameter and stored in the memory. The variable frequency lifting motor is controlled to drive the glass slide holder to rise vertically at a preset multi-level uniform speed gradient. The resistive strain sensor in the multi-source sensing and monitoring component is used to collect the signal spectral density and establish the mechanical background noise compensation spectrum. The variable frequency lifting motor is driven to control the slide holder to rise vertically with uniform acceleration. The real-time vertical acceleration values ​​output by the triaxial inertial space acceleration and angular velocity sensors in the multi-source sensing and monitoring components are read simultaneously to fit the feature mapping matrix of the acceleration-inertial impedance amplitude compensation coefficient. The static zero-point offset parameters are updated during the static gap before the slide holder is removed from the liquid cylinder each time.

6. The control method according to claim 4, characterized in that, The step of acquiring the dynamic tensile impedance electrical signal sequence and the spatial attitude dynamic electrical signal sequence during the lifting process and encapsulating them into a multi-dimensional temporal feature data frame includes: The variable frequency lifting motor is controlled to drive the slide holder to rise vertically and detach from the liquid surface at a first preset acceleration in the acceleration lifting section, and to drive the slide holder to move upward at a first preset speed in the constant speed running section. At the moment when the variable frequency lifting motor receives the start control command, the main control processor synchronously acquires the dynamic tensile impedance electrical signal sequence and the spatial attitude dynamic electrical signal sequence at a preset sampling rate. Based on the synchronization time reference corresponding to the preset sampling rate, the two synchronously acquired electrical signal sequences are time-aligned, and the aligned electrical signals are extracted using a sliding time window algorithm and encapsulated into the multi-dimensional time-series feature data frame. The physical prior constant of the reagent currently leaving the liquid cylinder is retrieved, and the physical prior constant is injected into the multidimensional time-series feature data frame as a feature weight coefficient.

7. The control method according to claim 4, characterized in that, The step of decoupling and outputting the absolute liquid adhesion mass of the liquid entrained on the slide holder includes: The dynamic stretching impedance electrical signal sequence in the multidimensional time-series feature data frame is subjected to discrete Fourier transform, and then transformed into an absolute quality feature spectrum with two dimensions in the time and frequency domains using the time window overlap and shift method. Read the mechanical background noise compensation spectrum, perform adaptive band-stop suppression on the mechanical resonance noise values ​​within the first preset frequency band in the absolute quality feature spectrum to filter out the mechanical resonance interference signal, and extract and retain the absolute liquid spectral energy components below the second preset frequency band threshold in the filtered absolute quality feature spectrum. Extract the real-time vertical acceleration vector and multiply it by the acceleration-inertial impedance amplitude compensation coefficient feature mapping matrix to obtain the inertial additional impedance stretching component. Then, subtract the inertial additional impedance stretching component from the retained absolute liquid spectral energy component in real time to physically peel off the gravitational stretching response and obtain the inertial compensated signal to be calculated. The inertial-compensated signal to be processed is input into a multi-path parallel characteristic bottleneck impedance calculation network. The slow gravitational field jump characteristics caused by the net mass of the reagent, the transient non-gravitational oscillation components caused by the undulation of the reagent surface, the sliding friction force of the robotic arm caused by axial micro-sliding, and the tangential impact reaction force generated by the air sweep pressure of the nozzle on the glass slide are extracted through four parallel physical branch channels. The absolute liquid wall mass is output through splicing and weighted average fusion decoupling.

8. The control method according to claim 4, characterized in that, The step of adaptively solving for a multi-dimensional pulse purging parameter matrix based on the absolute liquid wall adhesion mass using an unsupervised operating condition feature migration mechanism combined with the current physicochemical and operating condition data of the liquid cylinder, and driving the fluid purging execution component to perform the purging action, includes: Read the pre-stored calibration condition characteristic data and obtain the target reagent type identifier of the current liquid cylinder. Use the corresponding basic physical property parameter vector composed of reagent density, dynamic viscosity and surface tension coefficient as boundary constraint terms and input them into the unsupervised physical condition characteristic transfer adaptive control network. Using the unsupervised physicochemical condition feature migration adaptive control network, the pre-stored calibration condition features are migrated to the actual condition of the current liquid cylinder. By minimizing the physical field joint loss function that includes control deviation and time-varying physicochemical features of the reagent, the feature space reconstruction alignment between the calibration condition and the actual condition is achieved, and the control decision parameter vector of the current purging stage and the corresponding confidence score are output. A time-adaptive iterative mechanism is introduced, and the time consistency constraint in the physical field joint loss function is used to dynamically weight and correct the credibility score. When the corrected credibility score meets the preset credibility threshold, the actual motion trajectory feature corresponding to the control decision parameter vector is used as the update feature and injected back into the calibration condition feature data to realize the dynamic time-domain expansion of the calibration feature space. Based on the unsupervised physical and chemical condition feature transfer adaptive control network that has completed the feature space reconstruction and alignment, and combined with the absolute liquid wall mass, the control decision parameter vector is mapped and solved to generate a multi-dimensional pulse purging parameter matrix that conforms to the current physical characteristics of the liquid cylinder, which is used to drive the proportional pressure regulating valve and solenoid valve in the fluid purging execution component to perform sequential on / off actions.

9. The control method according to claim 8, characterized in that, The step of driving the proportional pressure regulating valve and solenoid valve in the fluid purging actuator to perform sequential on / off actions includes: Based on the absolute liquid wall mass output by real-time decoupling, the mass deviation value and change rate are calculated to determine the real-time state characteristics reflecting the dynamic sliding trend of the liquid film. Using the multidimensional pulse purge parameter matrix as a control reference, an event-triggered predictive execution unit is introduced to perform hierarchical calculation control in each sampling period based on the real-time state characteristics: when the real-time state characteristics meet the first channel triggering condition, the control signal in the multidimensional pulse purge parameter matrix of the previous period is sequentially retrieved and reused from the storage buffer configured in the event-triggered predictive execution unit; when the real-time state characteristics meet the second channel triggering condition, the time-domain predictive control calculation is activated to dynamically adjust the timing of the multidimensional pulse purge parameter matrix, update the storage buffer, and output the updated control sequence. The dual-channel upper-level safety limit coverage monitor in the event-triggered predictive execution unit performs online monitoring. When the absolute liquid wall mass reaches or falls below the safety residual set threshold, or when the number of consecutive multiplexing steps of the control signal in the storage buffer reaches the preset time domain limit, a zero bias voltage signal is forcibly output to the solenoid valve in the fluid purging execution component, and the solenoid valve is instantaneously shut off within a preset shutdown delay.

10. The closed-loop control method for suspended material transfer across cylinders according to claim 4, characterized in that, After the solenoid valve configured in the control fluid purging actuator is shut off to block the airflow, the following steps are also included: Control the operation of the horizontal transfer drive motor and the variable frequency lifting motor to move the glass slide holder from the current work position to the preset target work position. The target work position includes the next-level target liquid cylinder, drying chamber, temporary storage area or unloading position.

Images