A negative pressure adsorption type liquid drop recovery device for non-contact spraying

Abstract

The present application relates to the technical fields of semiconductor processing equipment automation control, and especially relates to a negative pressure adsorption type liquid drop recovery device for non-contact spraying, comprising a device and a control device, which collects absolute spatial coordinates and broadband shock wave signals through a position sensing assembly and a broadband detection assembly. A data acquisition module converts the coordinates into spatial curvature vectors; a distortion calculation module deduces the flow field separation singular point position and outputs the vortex choke index using fluid mechanics; a time delay estimation module calculates the transient resistance increment of the flexible air pipe and maps it into a time-space multi-dimensional dynamic delay matrix; a waveform optimization module generates a reverse compensation sound wave topology map to output the jet flow excitation parameters and the rotational speed adjustment value. A cooperative driving module controls the synchronous action of the variable frequency negative pressure source and the pulse excitation ring according to the time advance, uses high pressure micro air flow and the coanda effect to reshape the boundary layer velocity type distribution characteristics, eliminates the nonlinear air resistance mutation caused by the follow-up bending, and maintains the stability of the end suction.

Description

Technical Field

[0001] This invention relates to the field of automated control technology for semiconductor processing equipment, and in particular to a negative pressure adsorption droplet recovery device for non-contact spraying. Background Technology

[0002] The negative pressure adsorption droplet recovery device for non-contact spraying is an industrial equipment for collecting splashed pharmaceutical liquids. Non-contact spraying refers to a process in which the nozzle maintains a set spatial distance from the substrate and sprays liquid directly onto the substrate surface; due to the high-speed impact of the liquid on the substrate surface, tiny droplets inevitably bounce and splash. Negative pressure adsorption refers to using an air extraction device to create a pressure difference lower than the external atmospheric pressure in a closed environment. The higher external atmospheric pressure forces air carrying suspended droplets through a microporous suction nozzle into the annular negative pressure chamber. The droplets entering the chamber then merge into the continuous airflow to complete in-situ collection. Centralized negative pressure sources are usually fixedly installed in a stationary position outside the production line. Since the spraying operation components are in a continuously moving state, the flexible air duct responsible for transmitting airflow forms the physical channel connecting the fixed negative pressure source and the moving collection components.

[0003] Existing non-contact spray droplet recovery technologies suffer from the following technical challenges: specifically, the need to address the aerodynamic lag and dynamic instability of end-effector suction caused by nonlinear abrupt changes in air resistance due to the bending of flexible ducts. In high-precision wet etching processes on large-size glass substrates, the annular negative pressure chamber needs to move extensively along the glass substrate surface following the spray unit. This movement causes irregular bending and deformation of the flexible duct connected to the negative pressure source. This bending deformation alters the cross-sectional area and flow field distribution of the internal airflow channels, leading to abrupt changes in nonlinear air resistance within the duct. Conventional closed-loop control systems detect pressure deviations and issue adjustment commands. However, the negative pressure wave experiences severe aerodynamic lag during transmission through the nonlinear air resistance duct, preventing the control feedback loop from implementing real-time and accurate compensation. The dynamic instability of end-effector suction manifests as drastic fluctuations in suction at the micro-orifice nozzle. For example, when the spray unit moves rapidly from the center to the edge of the glass substrate, the duct bends sharply, generating significant resistance. The sudden drop in suction at the micro-orifice nozzle prevents the complete collection of splashed etching solution. After the control system detects insufficient suction, it issues a command to increase the suction power. Subsequently, the liquid spraying unit moves in the opposite direction, causing the air duct to flatten instantly. The sudden reduction in duct resistance triggers an overcurrent strong suction at the end. The overcurrent strong airflow severely disturbs and destroys the uniform chemical etching reaction state on the surface of the glass substrate. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a negative pressure adsorption-type droplet recovery device for non-contact spraying. This invention solves the technical problem that the aerodynamic pure lag in negative pressure feedback regulation and the dynamic instability of end suction caused by the nonlinear air resistance change caused by the flexible duct bending.

[0005] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows: This invention provides a negative pressure adsorption-type droplet recovery device for non-contact spraying, comprising an equipment unit and a control unit, wherein the control unit establishes a communication connection with the equipment unit; the equipment unit includes a spraying unit, an annular negative pressure chamber, a flexible duct, a variable frequency negative pressure source, a position sensing component, a pulse excitation ring, and a broadband detector component; the annular negative pressure chamber is disposed around the spraying unit and is connected to the beginning of the flexible duct, the end of the flexible duct is connected to the variable frequency negative pressure source, the pulse excitation ring is installed at the junction of the annular negative pressure chamber and the flexible duct, the position sensing component is disposed on the spraying unit, and the broadband detector component is disposed on the inner wall of the flexible duct; The control device includes: Data acquisition module: used to obtain absolute spatial coordinates through position sensing components, extract motion trajectory parameters from absolute spatial coordinates, and convert motion trajectory parameters into spatial curvature vectors; Distortion calculation module: used to receive spatial curvature vector and broadband shock wave signal obtained by broadband detection component, use spatial curvature vector and broadband shock wave signal to deduce the location of flow field separation singularity, and output vortex blockage index according to fluid kinetic energy dissipation rate at the location of flow field separation singularity. Time Delay Prediction Module: Used to receive the vortex congestion index, calculate the transient resistance increment caused by the inward contraction deformation of the flexible duct under negative pressure, and map the transient resistance increment into a spatiotemporal multidimensional dynamic delay matrix. Waveform optimization module: used to receive the spatiotemporal multidimensional dynamic delay matrix, extract the fluid oscillation parameters in the spatiotemporal multidimensional dynamic delay matrix to generate an anti-phase compensation acoustic wave topology map, and output the jet excitation parameters and speed adjustment value according to the anti-phase compensation acoustic wave topology map; Collaborative drive module: Used to receive jet excitation parameters and speed adjustment values, send speed adjustment electrical signals to the frequency conversion negative pressure source, and synchronously send pulse excitation action electrical signals to the pulse excitation ring.

[0006] Furthermore, the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention includes a data acquisition module comprising a coordinate filtering unit and a base frame conversion unit. The coordinate filtering unit obtains the absolute spatial coordinates, uses the Kalman filter algorithm to filter out the mechanical vibration noise in the absolute spatial coordinates, and outputs the noise-reduced coordinate data. The base frame conversion unit receives denoised coordinate data, constructs a three-dimensional active frame using the denoised coordinate data, extracts the motion tangent vector, normal vector, and subnormal vector, calculates the first derivative of the motion tangent vector with respect to the arc length parameter to obtain the bending curvature, calculates the first derivative of the subnormal vector with respect to the arc length parameter to obtain the spatial deflection, and outputs the bending curvature and spatial deflection as a spatial curvature vector including bending curvature and spatial deflection to the distortion calculation module.

[0007] Furthermore, the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention includes a distortion calculation module comprising a frequency domain decoupling unit and a partial differential calculation unit. The frequency domain decoupling unit receives the broadband shock wave signal, performs a fast Fourier transform on the broadband shock wave signal, and extracts the characteristic frequency physical quantities. The partial differential equation unit synchronously receives the spatial curvature vector and the characteristic frequency physical quantity, maps the spatial curvature vector into a centrifugal inertial force term, introduces the centrifugal inertial force term into the momentum conservation boundary condition of the Navier-Stokes partial differential equation, solves the in-flow velocity partial derivative approaching the inner wall of the flexible duct, marks the spatial coordinate point where the in-flow velocity partial derivative is equal to zero as the flow field separation singularity position, extracts the amplitude of the characteristic frequency physical quantity at the flow field separation singularity position, and uses the amplitude to calculate the fluid kinetic energy dissipation rate at the flow field separation singularity position to establish the vortex congestion index.

[0008] Furthermore, the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention includes a time delay prediction module comprising a viscoelastic observation unit and a matrix mapping unit. The viscoelastic observation unit receives the vortex congestion index, retrieves the preset polymer material relaxation time constant, sets the polymer material relaxation time constant as the calculus order of the fractional calculus operator, inputs the vortex congestion index as the excitation source into the fractional calculus operator, and uses the fractional calculus operator to calculate the transient resistance increment caused by the inward contraction deformation of the flexible duct under the action of the vortex congestion index. The matrix mapping unit receives the transient resistance increment, divides the transient resistance increment into a discretized grid along the axial length of the flexible duct, extracts the spatial location parameters of the discretized grid, and extends and transforms the single time dimension delay into a spatiotemporal multidimensional dynamic delay matrix that includes the spatial location parameters.

[0009] Furthermore, the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention includes a waveform optimization module comprising a waveform phase inversion unit and a parameter allocation unit. The waveform inversion unit receives the spatiotemporal multidimensional dynamic delay matrix, extracts the fluid oscillation amplitude distribution map and phase angle distribution matrix from the spatiotemporal multidimensional dynamic delay matrix as fluid oscillation parameters, increases the values ​​of all elements in the phase angle distribution matrix by 180 degrees, and reconstructs the phase angle distribution matrix after increasing the values ​​by 180 degrees with the fluid oscillation amplitude distribution map to obtain the anti-phase compensation acoustic wave topology map. The parameter allocation unit receives the anti-phase compensation acoustic wave topology diagram, uses Hilbert transform to analyze the anti-phase compensation acoustic wave topology diagram to obtain the instantaneous analytical signal, extracts the amplitude envelope extreme value of the instantaneous analytical signal, calculates the amplitude envelope extreme value as the jet excitation parameter of the driving pulse excitation ring, and simultaneously converts the amplitude envelope extreme value into the speed adjustment value to offset the macroscopic resistance of the pipeline.

[0010] Furthermore, in the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention, the jet excitation parameters calculated by the parameter allocation unit include the Strouhal locking frequency and the adaptive pulse width duty cycle. The parameter allocation unit uses the spatial curvature vector to calculate the pipe section contraction diameter of the flexible duct, sets the pipe section contraction diameter as the local airflow characteristic length, obtains the real-time air extraction volume of the variable frequency negative pressure source to calculate the mainstream time-averaged velocity, extracts the characteristic frequency physical quantity from the broadband shock wave signal, multiplies the characteristic frequency physical quantity by the local airflow characteristic length and divides it by the mainstream time-averaged velocity to obtain the Strouhal lock frequency.

[0011] Furthermore, the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention includes a phase advance compensation unit and a frequency conversion output unit in its collaborative drive module. The phase advance compensation unit receives the jet excitation parameters and the spatiotemporal multidimensional dynamic delay matrix, extracts the time lag variable corresponding to the current spatial position in the spatiotemporal multidimensional dynamic delay matrix, sets the time lag variable as the time advance, and sends the pulse excitation action electrical signal including the jet excitation parameters to the pulse excitation ring in advance according to the time advance. The frequency converter output unit receives the speed adjustment value and the time advance, and sends the speed adjustment electrical signal to the frequency converter negative voltage source in advance according to the time advance.

[0012] Furthermore, in the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention, the position sensing component includes a multi-axis grating ruler encoding component, which is installed on the side of the spraying unit to collect positioning pulse electrical signals and convert the positioning pulse electrical signals into absolute spatial coordinates. The broadband detection component includes a microelectromechanical high-frequency dynamic piezoelectric converter, which is flush-fitted into the bending stress stagnation point of the flexible duct inner wall, converting the collected flow field pressure pulsation into a continuous electrical waveform as a broadband shock wave signal.

[0013] Furthermore, in the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention, a piezoelectric ceramic micro-valve array is arranged around the pulse excitation ring, and the spray opening of the piezoelectric ceramic micro-valve array is inclined along the direction parallel to the tangent of the inner wall of the flexible air duct. The piezoelectric ceramic micro-valve array receives pulse-excited action signals to open and close instantaneously. It utilizes the Coanda effect to inject high-pressure micro-airflow tangentially into the boundary layer of the flexible duct. The high-pressure micro-airflow cuts off the energy return zone of the flow field separation bubbles generated inside the flexible duct.

[0014] Furthermore, the negative pressure adsorption droplet recovery device for non-contact spraying described in this invention has a centrifugal exhaust component inside the variable frequency negative pressure source. The variable frequency negative pressure source receives a speed adjustment electrical signal and changes the stator power supply frequency of the centrifugal exhaust component to change the air extraction volume. The changed air extraction volume is used in conjunction with the high-pressure micro-airflow injected by the piezoelectric ceramic micro-valve array to reshape the boundary layer velocity distribution characteristics at the flow field separation singularity position, so that the inner wall of the flexible duct is forcibly transformed from a non-slip state to an approximately slip state.

[0015] Beneficial effects of this invention: This invention provides a negative pressure adsorption-type droplet recovery device for non-contact spraying, which breaks through the physical response limit of conventional closed-loop feedback control in handling pipeline deformation. The control device uses a distortion calculation module to deeply integrate absolute spatial coordinates and broadband shock wave signals. The partial differential equation calculation unit introduces the centrifugal inertial force term into the Navier-Stokes partial differential equation to locate the singularity of the flow field separation and outputs the vortex congestion index. The control system transforms the passive monitoring of macroscopic air pressure into the active deduction of microscopic fluid dynamic state. The time delay prediction module uses the relaxation time constant of polymer materials and fractional-order calculus operators to construct a spatiotemporal multidimensional dynamic delay matrix. The waveform optimization module analyzes the anti-phase compensated acoustic wave topology to obtain the jet excitation parameters and speed adjustment value. The collaborative drive module generates a power signal in advance to the pulse excitation ring and the variable frequency negative pressure source according to the time advance, pre-compensating for the aerodynamic pure time delay caused by the transmission of negative pressure waves. The piezoelectric ceramic micro-valve array inside the pulse excitation ring opens and closes instantaneously, using the Coanda effect to inject high-pressure micro-airflow tangentially into the boundary layer of the flexible duct. High-pressure micro-airflow directly cuts off the energy return zone of the flow field separation bubbles. A variable-frequency negative pressure source synchronously changes the stator power supply frequency of the centrifugal exhaust assembly, altering the extraction volume. This altered extraction volume, combined with the high-pressure micro-airflow, reshapes the boundary layer velocity distribution at the flow field separation singularity, forcing the flexible duct's inner wall to transition from a non-slip state to a near-slip state. This rigorous physical intervention mechanism eliminates the nonlinear air resistance mutations caused by the flexible duct's bending at the source, completely overcoming the technical challenge of dynamic instability in the suction force at the micro-orifice nozzle tip. This creates a highly constant negative pressure flow field for chemical recovery in the high-precision wet etching process of large-size glass substrates. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on the accompanying drawings without creative effort.

[0017] Figure 1 This is a system architecture diagram of a negative pressure adsorption droplet recovery device for non-contact spraying according to the present invention. Detailed Implementation

[0018] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. The various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0019] Please see Figure 1 This invention provides a negative pressure adsorption-type droplet recovery device for non-contact spraying, comprising an equipment device and a control device, wherein the control device establishes a communication connection with the equipment device; the equipment device includes a spraying unit, an annular negative pressure chamber, a flexible duct, a variable frequency negative pressure source, a position sensing component, a pulse excitation ring, and a broadband detector component; the annular negative pressure chamber is disposed around the spraying unit and is connected to the beginning of the flexible duct, the end of the flexible duct is connected to the variable frequency negative pressure source, the pulse excitation ring is installed at the junction of the annular negative pressure chamber and the flexible duct, the position sensing component is disposed on the spraying unit, and the broadband detector component is disposed on the inner wall of the flexible duct; The control device includes: Data acquisition module: used to obtain absolute spatial coordinates through position sensing components, extract motion trajectory parameters from absolute spatial coordinates, and convert motion trajectory parameters into spatial curvature vectors; Distortion calculation module: used to receive spatial curvature vector and broadband shock wave signal obtained by broadband detection component, use spatial curvature vector and broadband shock wave signal to deduce the location of flow field separation singularity, and output vortex blockage index according to fluid kinetic energy dissipation rate at the location of flow field separation singularity. Time Delay Prediction Module: Used to receive the vortex congestion index, calculate the transient resistance increment caused by the inward contraction deformation of the flexible duct under negative pressure, and map the transient resistance increment into a spatiotemporal multidimensional dynamic delay matrix. Waveform optimization module: used to receive the spatiotemporal multidimensional dynamic delay matrix, extract the fluid oscillation parameters in the spatiotemporal multidimensional dynamic delay matrix to generate an anti-phase compensation acoustic wave topology map, and output the jet excitation parameters and speed adjustment value according to the anti-phase compensation acoustic wave topology map; Collaborative drive module: Used to receive jet excitation parameters and speed adjustment values, send speed adjustment electrical signals to the frequency conversion negative pressure source, and synchronously send pulse excitation action electrical signals to the pulse excitation ring.

[0020] In high-precision wet etching processes for large-size glass substrates, a negative pressure adsorption-type droplet recovery device for non-contact spraying is constructed according to a specific hardware topology. The spraying unit moves continuously above the glass substrate along a preset process path, with an annular negative pressure chamber tightly surrounding it. The annular negative pressure chamber moves synchronously with the spraying unit, using suction to capture tiny droplets of the splashed etching solution. The exhaust end of the annular negative pressure chamber connects to the beginning of a flexible duct, which acts as an airflow channel to transport the suspended droplets to a fixed area away from the production line. The end of the flexible duct connects to a variable frequency negative pressure source, which maintains a negative pressure below atmospheric pressure using an internally configured centrifugal exhaust component. A pulsed excitation ring is fitted at the physical junction of the annular negative pressure chamber and the flexible duct. A position sensing component is fixedly mounted on the movable electromechanical base of the spraying unit, and a broadband detector is flush-mounted against the inner wall of the flexible duct.

[0021] The control device establishes a low-level communication connection with the equipment and triggers the data processing logic of the data acquisition module. The position sensing component senses and outputs absolute spatial coordinates in real time, which encompass the length, width, and height spatial positioning values ​​in a three-dimensional physical coordinate system. The data acquisition module reads the absolute spatial coordinates under continuous timestamps and extracts the motion trajectory parameters from them. The motion trajectory parameters characterize the instantaneous displacement and directional deflection angle generated by the liquid spraying unit moving along the surface of the glass substrate. The data acquisition module performs differential mathematical operations on the motion trajectory parameters, converting them into a spatial curvature vector. The spatial curvature vector objectively quantifies the degree of geometric bending of the flexible duct as it moves with the liquid spraying unit. The data acquisition module sends the output spatial curvature vector to the next processing level along the internal bus.

[0022] The distortion calculation module receives the spatial curvature vector and performs low-level fluid dynamics calculations. A broadband detector synchronously acquires continuous electrical waveforms generated by pressure pulsations in the boundary layer inside the flexible duct. This waveform is then sent to the distortion calculation module as a broadband shock signal. The module combines the duct's physical deformation parameters reflected by the spatial curvature vector with the flow field pressure variation characteristics reflected by the broadband shock signal to deduce the location of the flow field separation singularity. This singularity represents the specific physical spatial coordinates of the localized turbulent vortices generated when the airflow detaches from the duct wall's attachment boundary. The module then assesses the state based on the fluid kinetic energy dissipation rate at the singularity location. This rate reflects the energy loss due to friction and collisions between airflow particles. The module outputs a vortex congestion index, quantifying the degree of duct blockage, based on the energy loss status.

[0023] The vortex congestion index is incorporated into the time delay prediction module for mathematical calculation of the pipeline's dynamic response delay. The time delay prediction module receives the vortex congestion index and uses it as an excitation source in the polymer material deformation observation algorithm. The module calculates the transient resistance increment caused by the inward contraction deformation of the flexible duct under negative pressure. This transient resistance increment reflects the additional nonlinear aerodynamic drag generated by the reduction in cross-sectional area of ​​the flexible duct under strong internal negative pressure suction and external atmospheric pressure. The time delay prediction module divides the transient resistance increment in a single time dimension into a discretized grid along the physical axial length of the flexible duct, mapping and extending the resistance change characteristics into a spatiotemporal multidimensional dynamic delay matrix. This spatiotemporal multidimensional dynamic delay matrix integrates the time variable values ​​of aerodynamic pure time delays at different spatial locations along the pipeline at specific time points.

[0024] The waveform optimization module receives the spatiotemporal multidimensional dynamic delay matrix and performs analytical calculations on the inverse compensation physical quantities. It extracts fluid oscillation parameters from the spatiotemporal multidimensional dynamic delay matrix, which encompass the amplitude and phase angle information of the pressure fluctuations within the pipe. The module reverses the phase angle of the fluid oscillation parameters by 180 degrees, generating an inverse-phase compensated acoustic wave topology that completely cancels out the original pressure fluctuations. Based on the inverse-phase compensated acoustic wave topology, the module decomposes the physical quantities and outputs the jet excitation parameters that drive the underlying micro-valve action and the speed adjustment value that changes the pumping power. The collaborative drive module receives the jet excitation parameters and the speed adjustment value, and sends a speed adjustment electrical signal to the variable frequency negative pressure source to adjust the stator power supply frequency and change the macroscopic pumping volume. Simultaneously, the collaborative drive module sends a pulse excitation action electrical signal to the pulse excitation ring. The jet excitation parameters drive the valve array at the junction to open and close instantaneously, injecting high-pressure micro-flow tangentially. The micro-jet, combined with macroscopic variable frequency suction, directly forces the reshaping of the boundary layer fluid operating state in the pipe wall flow field separation region.

[0025] In high-precision wet etching processes for large-size glass substrates, the coordinate filtering unit within the data acquisition module reads the absolute spatial coordinates from external input. These absolute spatial coordinates are mixed with high-frequency mechanical vibration noise transmitted from the moving electromechanical base of the spraying unit. The coordinate filtering unit uses the state prediction equation and state update equation in the Kalman filter algorithm to remove irregular high-frequency oscillation components and output denoised coordinate data. This denoised coordinate data represents the pure three-dimensional physical space positioning trajectory data after eliminating mechanical vibration interference. The frame conversion unit receives the denoised coordinate data and constructs a three-dimensional moving frame along the tangent, principal normal, and secondary normal directions of the spraying unit's moving trajectory. The frame conversion unit extracts the motion tangent vector, normal vector, and secondary normal vector from the three-dimensional moving frame. The frame conversion unit calculates the first derivative of the motion tangent vector with respect to the arc length parameter of the motion trajectory to obtain the bending curvature, which objectively reflects the degree of pipe bending. The frame conversion unit calculates the first derivative of the secondary normal vector with respect to the arc length parameter to obtain the spatial deflection, which reflects the degree of pipe twisting away from the oscillating plane. The base frame conversion unit outputs a combination of bending curvature and spatial deflection as a spatial curvature vector including bending curvature and spatial deflection, which is then sent to the distortion calculation module to perform flow field deformation deduction.

[0026] The frequency domain decoupling unit within the distortion calculation module continuously receives broadband shock wave signals. These signals carry complex acoustic frequency information generated by turbulent airflow in the duct wall boundary layer. The frequency domain decoupling unit performs a Fast Fourier Transform (FFT) on the broadband shock wave signals, converting the original time-domain waveform into a frequency-domain spectral distribution. From this spectral distribution, the unit identifies and extracts characteristic frequency physical quantities representing the frequency characteristics of local airflow vortex shedding. The partial differential equation (PDE) calculation unit simultaneously receives the spatial curvature vector and the characteristic frequency physical quantities, mapping the spatial curvature vector, representing macroscopic physical deformation, into a centrifugal inertial force term representing the forced motion of fluid particles within the duct. The PDE calculation unit incorporates this centrifugal inertial force term into the momentum conservation boundary conditions of the Navier-Stokes partial differential equations, solving for the downstream velocity partial derivative near the inner wall of the flexible duct. Fluid particles close to the inner wall of a flexible duct experience significant frictional resistance, leading to kinetic energy attenuation and even flow reversal. The partial differential equation (PDE) unit marks the spatial coordinate point where the partial derivative of the downstream velocity is zero as the singularity point where the airflow completely detaches from the duct wall. The PDE unit extracts the amplitude values ​​of characteristic frequency physical quantities at the singularity point and uses these amplitudes to calculate the fluid kinetic energy dissipation rate caused by friction and collision between fluid particles at the singularity point. This fluid kinetic energy dissipation rate is then used as the basis for establishing a vortex congestion index, representing the severity of airflow blockage in the duct.

[0027] The viscoelasticity observation unit within the time delay prediction module receives the vortex congestion index and participates in the mathematical derivation of the pipeline's dynamic response delay. The flexible duct wall material exhibits extremely pronounced viscoelastic physical properties, and the viscoelasticity observation unit retrieves the pre-entered relaxation time constant of the polymer material. This relaxation time constant objectively reflects the inherent sluggish time property of the polymer chain segments in the duct wall responding to externally applied stress. The viscoelasticity observation unit sets the polymer material relaxation time constant to the calculus order of a fractional-order calculus operator and inputs the vortex congestion index as an external aerodynamic excitation source into the fractional-order calculus operator model to perform nonlinear calculations. The fractional-order calculus operator derives the transient drag increment caused by the inward contraction deformation of the duct wall due to the abnormal flow field represented by the vortex congestion index under the influence of the abnormal flow field. The matrix mapping unit receives the transient drag increment data and divides it into multiple interconnected discretized grids along the physical axial extension of the flexible duct. The matrix mapping unit extracts the absolute spatial position parameters corresponding to all discretized grids, binds the pure delay values, which were originally in a single time dimension, with the spatial position parameters, and performs multi-dimensional coordinate transformation, outputting a spatiotemporal multidimensional dynamic delay matrix that includes the spatial position parameters. This spatiotemporal multidimensional dynamic delay matrix meticulously records the associated attribute data of aerodynamic response hysteresis at each physical cross-section of the flexible duct.

[0028] The waveform inversion unit within the waveform optimization module receives the spatiotemporal multidimensional dynamic delay matrix and performs analytical calculations on the inverse physical quantities. The waveform inversion unit extracts from the spatiotemporal multidimensional dynamic delay matrix a fluid oscillation amplitude distribution map reflecting the intensity of airflow fluctuations and a phase angle distribution matrix reflecting the alternating time pattern of airflow fluctuations. The fluid oscillation amplitude distribution map and the phase angle distribution matrix together constitute the fluid oscillation parameters to be processed. The waveform inversion unit increments the values ​​of all elements in the phase angle distribution matrix by 180 degrees to perform phase inversion. The waveform inversion unit then reconstructs and combines the phase angle distribution matrix (after incrementing the values ​​by 180 degrees) with the fluid oscillation amplitude distribution map at the mathematical attribute level, resulting in an inverse-phase compensated acoustic wave topology map used to guide the underlying hardware in performing inverse-phase physical cancellation. The parameter allocation unit receives the inverse-phase compensated acoustic wave topology map and performs orthogonal analytical operations on it using the Hilbert transform algorithm, obtaining an instantaneous analytical signal carrying instantaneous amplitude and instantaneous frequency physical information. The parameter allocation unit extracts the highest energy contour boundary data corresponding to the instantaneous analytical signal as the amplitude envelope extremum. The parameter allocation unit performs conversion and calculation of the amplitude envelope extreme values ​​to obtain the jet excitation parameters that drive the bottom micro valve hardware action. Simultaneously, the parameter allocation unit converts and calculates the amplitude envelope extreme values ​​into speed adjustment values ​​that offset the macroscopic airflow friction resistance of the entire pipeline and sends them to the collaborative drive module.

[0029] In high-precision wet etching processes for large-size glass substrates, the parameter allocation unit within the waveform optimization module receives the previously calculated inverse-phase compensated acoustic topology data and performs specific physical quantity calculations for the jet excitation parameters. The jet excitation parameters calculated by the parameter allocation unit include two underlying control physical quantities: the Strouhal locking frequency and the adaptive pulse width duty cycle. The parameter allocation unit reads the three-dimensional geometric bending degree recorded in the spatial curvature vector and uses this vector to calculate the pipe cross-sectional shrinkage diameter after compression deformation under the current physical bending state. The pipe cross-sectional shrinkage diameter objectively reflects the minimum flow space span after a circular duct is mechanically compressed into an elliptical cross-section. The parameter allocation unit sets the pipe cross-sectional shrinkage diameter as the local airflow characteristic length describing the degree of flow field boundary reduction. Simultaneously, the parameter allocation unit acquires the real-time air extraction volume value fed back by the variable frequency negative pressure source sensor and, combined with the standard inner diameter data of the flexible duct, derives and calculates the mainstream time-averaged velocity for stable fluid flow inside the duct. The parameter allocation unit extracts characteristic frequency physical quantities representing the periodic properties of local turbulent vortex shedding from the broadband shock wave signal transmitted by the broadband detector component. Using dimensionless derivation logic from fluid mechanics, the unit multiplies the characteristic frequency physical quantity by the local airflow characteristic length and divides it by the mainstream time-averaged velocity to derive the Strouhal locking frequency that dynamically matches the current aerodynamic congestion state of the pipeline.

[0030] The collaborative drive module is internally divided into a phase advance compensation unit that executes advance prediction logic and a frequency conversion output unit. The phase advance compensation unit receives jet excitation parameters from the parameter allocation unit and a spatiotemporal multidimensional dynamic delay matrix generated by the time delay estimation module. The spatiotemporal multidimensional dynamic delay matrix stores a three-dimensional mapping relationship between the length coordinates of different physical axes along the pipeline and the aerodynamic response lag time. The phase advance compensation unit inputs the absolute spatial coordinate position parameters under the current calculation cycle and extracts the time lag variable corresponding to the current spatial position from the spatiotemporal multidimensional dynamic delay matrix. The time lag variable objectively quantifies the pure physical delay time span required for the transmission of air pressure fluctuations within a long-distance flexible duct. The phase advance compensation unit directly sets the extracted time lag variable as the time advance of the closed-loop control loop. According to the time scale set by the time advance, the phase advance compensation unit sends a pulse excitation action electrical signal, including the jet excitation parameters, to the pulse excitation loop at the pipeline junction in advance. The frequency converter output unit synchronously receives the calculated speed adjustment value and time advance, and sends the speed adjustment electrical signal to the frequency converter negative pressure source at the end of the system in advance according to the time advance scale specified by the time advance.

[0031] In terms of the physical detection hardware layout, the position sensing component is specifically configured as an industrial-grade multi-axis grating ruler encoding component. This component is securely mounted on a linear metal moving part on the side of the spray unit. The photoelectric scanning element inside the multi-axis grating ruler encoding component dynamically reads the changes in the lines of the glass reference grating, acquiring high-frequency micron-level positioning pulse electrical signals. The data processing chip inside the multi-axis grating ruler encoding component converts the positioning pulse electrical signals into absolute spatial coordinate values ​​in a three-dimensional physical coordinate system. In terms of the flow field feature detection hardware layout, the broadband detection component is specifically configured as a microelectromechanical high-frequency dynamic piezoelectric converter. This converter is tightly arranged at the bending stress points of the flexible duct inner wall using a flush-fit installation process. This flush-fit physical installation method avoids secondary airflow blockage and flow field disturbance problems caused by the sensor probe protruding from the duct wall. The microelectromechanical high-frequency dynamic piezoelectric converter senses the flow field pressure pulsation phenomenon occurring at the bending stress stagnation point. The crystal inside the converter generates a piezoelectric effect under stress, and converts the collected flow field pressure pulsation into a continuous electrical waveform voltage signal as a broadband shock wave signal output to the control bus.

[0032] Inside the pulse excitation ring, a dense array of piezoelectric ceramic micro-valve arrays is arranged around the circumferential pipe wall. The independent injection openings of the piezoelectric ceramic micro-valve array are not perpendicular to the central axis of the pipe; instead, they are inclined along the tangent to the inner wall of the flexible duct, according to a set hydrodynamic guiding angle. The piezoelectric ceramic micro-valve array receives an advanced pulse excitation signal, driving the internal components to open and close instantaneously. This high-frequency mechanical action of instantaneous opening and closing releases a micro-airflow exhibiting an extremely high-frequency sub-pulse state. The high-pressure micro-airflow injected tangentially creates a local low-pressure boundary in the jet outlet region. The fluid in the outer boundary layer, under the influence of the pressure difference, moves towards the jet center and adheres tightly to the pipe wall surface, flowing at high speed, triggering the typical Coanda effect in fluid mechanics. Utilizing the Coanda effect, the piezoelectric ceramic micro-valve array continuously injects high-pressure micro-airflow tangentially into the boundary layer of the flexible duct. The injected high-pressure micro-airflow exhibits an extremely thin, high-speed laminar flow boundary state, using the physical kinetic energy of the high-pressure micro-airflow to forcibly cut off the energy supply and return zone of the flow field separation bubbles generated within the flexible duct.

[0033] The variable frequency negative pressure source is internally equipped with a centrifugal exhaust assembly for generating macroscopic suction and traction airflow. The variable frequency negative pressure source receives a speed adjustment electrical signal from the co-drive module via a communication cable. The internal power variable frequency drive controller of the variable frequency negative pressure source changes the stator power supply frequency output to the drive motor inside the centrifugal exhaust assembly. The dynamic increase or decrease of the stator power supply frequency directly changes the rotational angular velocity of the centrifugal fan impeller. The control system dynamically changes the overall volumetric air extraction rate of the pipeline by altering the stator power supply frequency of the centrifugal exhaust assembly. The entire device utilizes the altered air extraction rate in conjunction with a piezoelectric ceramic micro-valve array to inject high-pressure micro-airflow tangentially using the Coanda effect, constructing a dual physical intervention mechanism that combines macroscopic suction and traction with microscopic boundary layer kinetic energy blowing. This dual physical intervention mechanism directly acts on regions where the flow field is severely deteriorated, forcibly reshaping the boundary layer velocity distribution characteristics at the flow field separation singularity. The physical reshaping of velocity distribution characteristics allows fluid micro-elements near the bottom of the pipe wall to regain sufficient kinetic energy in the downstream direction, completely breaking the near-wall stagnation state where the flow velocity is zero, and forcing the inner wall of the flexible duct from its inherent non-slip state to an approximate slip state.

[0034] The data acquisition module includes a coordinate filtering unit and a base transformation unit. The coordinate filtering unit acquires absolute spatial coordinates and uses a Kalman filter algorithm to filter out mechanical vibration noise from the absolute spatial coordinates, outputting denoised coordinate data. The specific data processing path for using the Kalman filter algorithm to filter out mechanical vibration noise from the absolute spatial coordinates involves establishing the system state prediction equation and the system state update equation. The system state prediction equation is:

[0035] In the formula, Represents the prior state estimate for the current computation step; Represents the state transition matrix; This represents the optimal state estimate from the previous calculation step; Represents the control input matrix; This represents the control vector for the current computation step. The system state update equation is:

[0036] In the formula, This represents the optimal state estimate obtained in the current calculation step and output as the denoised coordinate data. Represents the prior state estimate for the current computation step; Represents the Kalman gain matrix; This represents the absolute spatial coordinates measured by the position sensing component at the current moment; The observation matrix is ​​represented by the following matrix. In the application scenario of high-precision wet etching process on large-size glass substrates, combined with the physical kinematic characteristics of the spraying unit, the state transition matrix A is configured as a 9x9 constant matrix including the three-dimensional position components, three-dimensional velocity components, and three-dimensional acceleration components of the spraying unit. The control input matrix B is configured as a 9x3 transformation matrix representing the conversion coefficient from the driving torque of the electromechanical base motor to the three-dimensional spatial displacement. The observation matrix H is configured as a 3x9 diagonal matrix mapping the physical parameters of absolute spatial coordinates to the measurement coordinate system of the multi-axis grating ruler encoding component. The control vector Uk is configured as the sequence of three-dimensional control voltage signals input to the moving motor of the spraying unit in the current calculation cycle. The above data dimension settings and the definition of the physical meaning of the matrix parameters establish the data input-output mapping between the physical hardware of the actual movement trajectory of the spraying unit and the Kalman filter mathematical model, and establish the intrinsic correlation between the algorithm and the specific mechanical motion scenario.

[0037] The base frame conversion unit receives denoised coordinate data, constructs a 3D moving frame using the denoised coordinate data, extracts the motion tangent vector, normal vector, and binormal vector, calculates the first derivative of the motion tangent vector with respect to the arc length parameter to obtain the bending curvature, and calculates the first derivative of the binormal vector with respect to the arc length parameter to obtain the spatial deflection. The bending curvature and spatial deflection are combined and output as a spatial curvature vector including both bending curvature and spatial deflection to the distortion calculation module. The specific calculation formula for obtaining the bending curvature by calculating the first derivative of the motion tangent vector with respect to the arc length parameter is as follows:

[0038] In the formula, Represents the curvature of the bend; Represents the motion tangent vector in a three-dimensional active frame; The arc length parameter representing the trajectory of motion; The absolute value of the first derivative of the tangent vector with respect to the arc length parameter represents the spatial torsion. The specific formula for calculating the spatial torsion is obtained by taking the first derivative of the binormal vector with respect to the arc length parameter:

[0039] In the formula, Represents spatial torsion; Represents the normal vector in a three-dimensional active frame; Represents the secondary normal vector in a three-dimensional active frame; The arc length parameter representing the trajectory of motion; This represents the first-order derivative of the binormal vector with respect to the arc length parameter.

[0040] The distortion calculation module includes a frequency domain decoupling unit and a partial differential equation (PDE) calculation unit. The frequency domain decoupling unit receives the broadband shock wave signal, performs a fast Fourier transform on it, and extracts the characteristic frequency physical quantities. The PDE calculation unit simultaneously receives the spatial curvature vector and the characteristic frequency physical quantities, maps the spatial curvature vector to a centrifugal inertial force term, introduces this term into the momentum conservation boundary conditions of the Navier-Stokes partial differential equations, solves for the downstream velocity partial derivative approaching the inner wall of the flexible duct, and marks the spatial coordinates where the downstream velocity partial derivative equals 0 as the location of the flow field separation singularity. The specific boundary layer separation determination equation for marking the location of the flow field separation singularity is as follows:

[0041] In the formula, The partial derivative of the downstream velocity near the inner wall of the flexible duct; Represents the downstream velocity component within the boundary layer of the flexible duct; The variable representing the normal absolute distance perpendicular to the inner wall surface of the flexible duct; This represents the physical boundary of the inner wall surface of the flexible duct. The partial differential equation unit extracts the amplitude of the characteristic frequency physical quantity at the flow field separation singularity location, and uses the amplitude to calculate the fluid kinetic energy dissipation rate at the flow field separation singularity location to establish the vortex choking index [cite:57]. The specific mathematical formula for calculating the fluid kinetic energy dissipation rate at the flow field separation singularity location is as follows:

[0042] In the formula, This represents the fluid kinetic energy dissipation rate at the location of the flow field separation singularity. The kinematic viscosity coefficient representing the fluid inside the pipe; The fluid velocity component represents the first direction in three-dimensional space; Represents the fluid velocity component in the second direction of three-dimensional space; Represents the spatial position coordinates in the first direction of three-dimensional space; This represents the spatial coordinates of the second direction in three-dimensional space. For the multiphase flow field state where the chemical droplets and air are mixed inside the negative pressure adsorption droplet recovery device, the kinematic viscosity coefficient ν cannot be the standard air viscosity. Instead, it is configured as an equivalent mixed kinematic viscosity parameter that integrates the standard air kinematic viscosity and the concentration ratio of the etching chemical droplets. The input data in the fluid kinetic energy dissipation rate calculation not only includes the spatial coordinates and fluid velocity components, but also uses the equivalent mixed kinematic viscosity parameter as a core input parameter, which, along with the flexible duct bending physical shape, acts on the momentum conservation boundary conditions of the Navier-Stokes partial differential equations. The vortex congestion index output by the partial differential calculation unit, as a single-dimensional floating-point value reflecting the degree of energy attenuation in the flow field, objectively maps the turbulent shedding state of the complex multiphase flow under nonlinear physical boundaries, clarifying the data interaction logic between the fluid dynamics calculation model and the specific processing environment of wet etching.

[0043] The time delay prediction module includes a viscoelastic observation unit and a matrix mapping unit. The viscoelastic observation unit receives the vortex congestion index, retrieves the preset polymer material relaxation time constant, sets the polymer material relaxation time constant as the calculus order of the fractional-order calculus operator, and uses the vortex congestion index as the excitation source input to the fractional-order calculus operator. The fractional-order calculus operator is then used to calculate the transient drag increment caused by the inward contraction deformation of the flexible duct under the action of the vortex congestion index. The specific fractional-order viscoelastic constitutive equation for calculating the transient drag increment is as follows:

[0044] In the formula, Represents the increase in transient resistance; The geometric damping constant of the duct wall representing the flexible duct material; Represents the relaxation time constant of polymer materials; This represents the calculus order defined by the relaxation time constant mapping of the polymer material; The order is represented as Fractional calculus operators; The vortex congestion index is represented as a function of time variables. Representing the time-space physical variable. The matrix mapping unit receives the transient drag increment, divides the transient drag increment into a discretized grid along the axial length of the flexible duct, extracts the spatial position parameters of the discretized grid, and extends and converts the single-time dimension delay into a spatiotemporal multidimensional dynamic delay matrix including the spatial position parameters. The partial differential equation calculation unit extracts the amplitude of the characteristic frequency physical quantity of 125 Hz at the calibration coordinate point and substitutes it into the energy dissipation formula to calculate that the fluid kinetic energy dissipation rate at the flow field separation singularity is 1.5 watts per kilogram, which is then converted into a dimensionless vortex congestion index of 8.2. The viscoelastic observation unit retrieves the polymer material relaxation time constant set for the flexible duct's polymer material as 0.6 seconds, inputs 0.6 seconds into the fractional-order viscoelastic constitutive equation as the calculus order of the fractional-order calculus operator, and derives that the transient drag increment generated by the flexible duct under the action of the vortex congestion index of 8.2 is 350 Pascals.

[0045] The waveform optimization module includes a waveform inversion unit and a parameter allocation unit. The waveform inversion unit receives the spatiotemporal multidimensional dynamic delay matrix, extracts the fluid oscillation amplitude distribution map and phase angle distribution matrix from the spatiotemporal multidimensional dynamic delay matrix as fluid oscillation parameters, increases the values ​​of all elements in the phase angle distribution matrix by 180 degrees, and reconstructs the phase angle distribution matrix with the fluid oscillation amplitude distribution map to obtain the inverse-phase compensated acoustic wave topology map. The parameter allocation unit receives the inverse-phase compensated acoustic wave topology map, uses Hilbert transform to analyze the inverse-phase compensated acoustic wave topology map to obtain the instantaneous analytical signal, extracts the amplitude envelope extrema of the instantaneous analytical signal, calculates the amplitude envelope extrema as the jet excitation parameters for driving the pulse excitation ring, and simultaneously converts the amplitude envelope extrema into speed regulation values ​​to offset the macroscopic resistance of the pipeline. The specific integral equation for obtaining the instantaneous analytical signal using Hilbert transform analysis of the inverse-phase compensated acoustic wave topology map is as follows:

[0046] In the formula, This represents the instantaneous analytical signal obtained after computation and processing; The representative is an inversely compensated acoustic topology data sequence that is a function of time variables; Represents the physical parameter of continuous integration operation time; The physical parameter representing the current observation time; This represents the mathematical constant pi. The jet excitation parameters calculated by the parameter allocation unit include the Strouhal lock-in frequency and the adaptive pulse width duty cycle. The parameter allocation unit uses the spatial curvature vector to calculate the duct cross-sectional contraction diameter of the flexible duct, sets the duct cross-sectional contraction diameter as the local airflow characteristic length, obtains the real-time pumping volume of the variable frequency negative pressure source to calculate the mainstream time-averaged velocity, extracts the characteristic frequency physical quantity from the broadband shock wave signal, multiplies the characteristic frequency physical quantity by the local airflow characteristic length and divides it by the mainstream time-averaged velocity to obtain the Strouhal lock-in frequency. In the step of extracting the amplitude envelope extremum of the instantaneous analytical signal, the parameter allocation unit sets a continuously sliding judgment window in the time dimension. The time span width of the sliding judgment window is dynamically adjusted according to the broadband shock wave signal input parameters, and the time span width is set to three to five times the complete cycle corresponding to the characteristic frequency physical quantity. By comparing the envelope peak data within the sliding judgment window, the parameter allocation unit simultaneously introduces a preset air pressure disturbance reference amplitude threshold as a filtering condition. Pseudo-extreme data with amplitude values ​​below the air pressure disturbance reference amplitude threshold are filtered out by the algorithm, while the highest energy profile data with amplitude values ​​above or equal to the air pressure disturbance reference amplitude threshold are extracted and output as amplitude envelope extreme values. The rules for setting the judgment window and establishing the threshold filtering conditions clarify the underlying digital signal processing logic of the waveform optimization model for separating real aerodynamic congestion characteristics from complex acoustic wave data sequences. The parameter allocation unit sets the physical reference value range of the air pressure disturbance reference amplitude threshold to between 50 Pascals and 150 Pascals. The parameter allocation unit continuously collects the fluid oscillation amplitude within a 10-second time window when the equipment is in a straight, bend-free operating state. The parameter allocation unit calculates the average value of the fluid oscillation amplitude under the straight, bend-free operating state, multiplies the average value by a safety margin factor of 1.5, and obtains the dynamically changing air pressure disturbance reference amplitude threshold. For example, if the parameter allocation unit measures the average fluid oscillation amplitude under straight-line operation without bends to be 60 Pascals, it multiplies this 60 Pascals by a safety margin of 1.5 to calculate the pressure disturbance reference amplitude threshold for the current operating batch as 90 Pascals. The parameter allocation unit directly filters out pseudo-extreme data with amplitude values ​​below 90 Pascals and simultaneously extracts the highest energy profile data with amplitude values ​​greater than or equal to 90 Pascals as the actual control parameter output.

[0047] The specific mathematical formula for calculating the Strauhall locking frequency is as follows:

[0048] In the formula, This represents the Strauhal locking frequency that needs to be solved; This represents the characteristic frequency physical quantity extracted from the broadband shock wave signal; Represents the local airflow characteristic length defined by the pipe cross-sectional diameter contraction; This represents the mainstream hourly average velocity calculated from the real-time air extraction volume of the variable frequency negative pressure source. The collaborative drive module extracts the time lag variable corresponding to the current spatial location and sets it as the time advance, controlling the pulse excitation ring and the variable frequency negative pressure source to perform physical intervention actions in advance according to the time advance.

[0049] To verify the feasibility of the data processing model and system equations, a complete data flow path including specific numerical values ​​was established for an embodiment. The data acquisition module obtains the absolute spatial coordinates measured by the injection unit at a specific processing moment. High-frequency vibration noise is removed using the system state prediction equation and system state update equation, and denoised coordinate data including coordinate points is calculated. Using the denoised coordinate data, the bending curvature representing the pipe bending state is calculated to be 0.05 per millimeter, and the spatial deflection representing the pipe twisting state is calculated to be 0.02 per millimeter. These are combined and packaged to output a spatial curvature vector. The frequency domain decoupling unit performs a fast Fourier transform calculation on the broadband shock wave signal, extracting the characteristic frequency physical quantity of 125 Hz included in the high-frequency pulsations of the pipe wall flow field. The partial differential equation (PDE) unit incorporates the centrifugal inertial force term into the momentum conservation boundary conditions of the Navier-Stokes partial differential equations. Calculations show that the partial derivative of the downstream velocity near the inner wall of the flexible duct is exactly zero at the coordinate point where the X-axis is 200 mm, the Y-axis is 150 mm, and the Z-axis is 50 mm. This coordinate point is designated as the flow field separation singularity. The PDE unit extracts the amplitude of the characteristic frequency physical quantity at 125 Hz at the designated coordinate point and substitutes it into the energy dissipation formula. The calculated fluid kinetic energy dissipation rate at the flow field separation singularity is 1.5 watts per kilogram. This 1.5 watts per kilogram is converted to a dimensionless vortex congestion index of 8.2. The viscoelastic observation unit retrieved the polymer material relaxation time constant set to 0.6 seconds for the flexible duct polymer material. The 0.6 seconds was input into the fractional-order viscoelastic constitutive equation as the calculus order of the fractional-order calculus operator, and the transient drag increment generated by the flexible duct under the action of vortex congestion index 8.2 was deduced to be 350 Pascals.

[0050] The matrix mapping unit converts the 350 Pascal extension into a spatiotemporal multidimensional dynamic delay matrix, matching and extracting the time lag variable of the current spatial position as 45 milliseconds, and directly setting 45 milliseconds as the underlying time advance of the system. The waveform inversion unit and parameter allocation unit perform mathematical analysis steps to obtain the local airflow characteristic length of the flexible duct as 0.08 meters, and calculate the mainstream average velocity as 20 meters per second based on the real-time air extraction volume of 500 cubic meters per hour from the variable frequency negative pressure source. The parameter allocation unit multiplies the 125 Hz characteristic frequency physical quantity by the local airflow characteristic length of 0.08 meters, divides it by the mainstream average velocity of 20 meters per second, and accurately calculates the Strouhal lock frequency used to drive the airflow pulse as 0.5. The collaborative drive module receives the Strouhal lock frequency of 0.5, and according to the standard set by the 45 millisecond time advance, sends the pulse excitation action electrical signal to the pulse excitation loop 45 milliseconds in advance, and simultaneously sends the speed adjustment electrical signal to the variable frequency negative pressure source 45 milliseconds in advance, achieving the goal of closed-loop control of aerodynamic parameters with highly adaptive characteristics. The data processing path within the control device for constructing the polymer material deformation observation model involves executing a dynamic thermomechanical analysis (MMA) test. The data processing system controls an external MMA instrument to apply alternating sinusoidal stress with a frequency ranging from 1 Hz to 100 Hz to the polyurethane pipe wall material sample of the flexible duct. Simultaneously, the MMA instrument collects storage modulus and loss modulus data of the polyurethane pipe wall material sample within the process environment temperature range of 25°C to 40°C. The data processing system divides the loss modulus data by the storage modulus data to obtain the loss tangent value. Based on the peak width of the loss tangent value distribution curve, the system calculates the pipe wall geometric damping constant. The system extracts the temperature and frequency coordinates corresponding to the peak value of the loss tangent value distribution curve and calculates the polymer material relaxation time constant using the time-temperature equivalence mathematical principle. The data processing system then stores the pipe wall geometric damping constant and the polymer material relaxation time constant in the underlying memory of the viscoelastic observation unit, completing the task of constructing the underlying algorithm for the polymer material deformation observation model.

[0051] Embodiment 1 of this invention: In the application scenario of high-precision wet etching process for large-size glass substrates, Embodiment 1 provides a bottom-level hardware linkage and data mapping scheme for a negative pressure adsorption droplet recovery device for non-contact spraying. When the production line processes 8.5-generation glass substrates, the spraying unit performs reciprocating spraying operations along the surface of the glass substrate at a moving speed of 500 mm / s. A multi-axis grating ruler encoding component installed on the side of the metal linear moving guide rail of the spraying unit uses photoelectric scanning elements to follow and read the change state of the grating lines, and high-frequency acquires the positioning pulse electrical signal of the electromechanical base. The data acquisition module inside the control device receives the positioning pulse electrical signal, calls the coordinate filtering unit to use the state prediction equation and state update equation in the Kalman filter algorithm to filter out the irregular high-frequency oscillation components caused by mechanical vibration, and outputs pure absolute spatial coordinate data. The base frame conversion unit constructs a three-dimensional active frame based on absolute spatial coordinate data. It calculates the first derivative of the arc length parameter of the motion trajectory along the motion tangent vector and the binormal vector, respectively, to objectively quantify the bending curvature and spatial deflection of the current physical bending state of the flexible duct. The bending curvature and spatial deflection are then combined and packaged into a spatial curvature vector and sent to the next computational level. Simultaneously, a microelectromechanical high-frequency dynamic piezoelectric converter, flush with the bending stress stagnation point on the inner wall of the flexible duct, synchronously senses the internal flow field pressure pulsation. Utilizing the crystal deformation stress characteristics, the microelectromechanical high-frequency dynamic piezoelectric converter converts the flow field pressure pulsation into a continuous electrical waveform, outputting a broadband shock wave signal which is then processed by the distortion calculation module for physical quantity decoupling calculation.

[0052] Embodiment 2 of this invention: Building upon the data mapping foundation established in Embodiment 1, Embodiment 2 focuses on illustrating the flow field state deduction and time-delay observation logic triggered by the control device when handling the large-scale span movement of a 10.5-generation glass substrate. The frequency domain decoupling unit within the distortion calculation module performs a Fast Fourier Transform operation on the received broadband shock wave signal, extracting characteristic frequency physical quantities characterizing the local airflow turbulence state. The partial differential equation calculation unit maps the spatial curvature vector into a centrifugal inertial force term representing the forced motion of fluid particles within the pipe. This centrifugal inertial force term is substituted into the momentum conservation boundary conditions of the Navier-Stokes partial differential equations, and the downstream velocity partial derivative is solved along the direction from the inner wall of the pipe towards the central axis. The partial differential equation (PDA) unit marks the three-dimensional spatial coordinates where the partial derivative of the downstream velocity is equal to 0 as the singularity of the flow field separation where the airflow detaches from the pipe wall. Combining the amplitude of characteristic frequency physical quantities at the singularity, it solves for the fluid kinetic energy dissipation rate caused by the friction and collision of fluid particles, outputting the vortex congestion index, which quantifies the aerodynamic blockage state of the pipeline. The time delay prediction module retrieves the 0.85-second relaxation time constant of the polymer material corresponding to the flexible duct, directly setting 0.85 seconds as the calculus order of the fractional calculus operator, and derives the transient drag increment caused by the inward contraction deformation of the flexible duct wall under the abnormal aerodynamic forces represented by the vortex congestion index. The matrix mapping unit binds the grid coordinates discretized along the axial length of the pipeline with the transient resistance increment in a high-dimensional data manner, and establishes a spatiotemporal multidimensional dynamic delay matrix that records the aerodynamic response lag time parameters at different physical locations. The waveform optimization module then uniformly increases the element values ​​of the phase angle distribution matrix inside the spatiotemporal multidimensional dynamic delay matrix by 180 degrees, and reconstructs the anti-phase compensation acoustic wave topology map used to offset the original air pressure fluctuations. The highest energy profile corresponding to the instantaneous analytical signal is extracted using Hilbert transform to obtain the amplitude envelope extremum.

[0053] Embodiment 3 of this invention: Based on the inverse physical topology model constructed in Embodiment 2, Embodiment 3 fully discloses the underlying execution process of physical intervention parameter allocation and dual-end collaborative drive. The parameter allocation unit deduces the pipe cross-section contraction diameter after the flexible duct is compressed into an elliptical cross-section based on the spatial curvature vector. It sets the pipe cross-section contraction diameter as the local airflow characteristic length describing the degree of flow field boundary reduction, and calculates the mainstream hourly velocity inside the pipe by combining the real-time volumetric air extraction rate of the variable frequency negative pressure source at the current operating moment. The parameter allocation unit multiplies the extracted characteristic frequency physical quantity by the local airflow characteristic length and divides it by the mainstream hourly velocity to derive the Strouhal locking frequency, which perfectly matches the shedding period of the turbulent vortex inside the pipe. The phase advance compensation unit in the collaborative drive module extracts the time lag variable corresponding to the current spatial position from the spatiotemporal multidimensional dynamic delay matrix and directly sets the time lag variable as the time advance of the system action. The variable frequency output unit, based on a 120-millisecond physical time span set by the time advance, sends a speed adjustment electrical signal to the external variable frequency negative pressure source of the production line in advance, adjusting the stator power supply frequency of the centrifugal exhaust component to change the macroscopic air extraction volume. The piezoelectric ceramic micro-valve array arranged inside the pulse excitation ring opens instantaneously under the drive of the pre-sent pulse excitation action signal, and, according to the action frequency limited by the Strauhall locking frequency, sprays high-pressure micro-airflow at an angle parallel to the tangential direction of the flexible duct's inner wall. During the tangential injection process, the high-pressure micro-airflow induces the Coanda effect, causing the micro-airflow to adhere tightly to the boundary layer of the flexible duct and flow at high speed. Utilizing the physical kinetic energy carried by the high-pressure micro-airflow, it cuts off the energy return zone of the flow field separation bubbles. Combined with the macroscopic suction action of the variable frequency negative pressure source after changing the fixed frequency parameters, it forcibly reshapes the boundary layer velocity distribution characteristics at the flow field separation singularity position, causing the inner wall of the flexible duct to forcibly change from a non-slip state to an approximately slipping state.

[0054] To verify the effectiveness of the dual physical intervention mechanism in stabilizing end suction, this invention conducted a comparative test experiment on the stability of the negative pressure flow field during droplet recovery. The experiment involved a spray unit moving at 800 mm / s on an 8.5-generation glass substrate surface, performing a trajectory-changing scanning motion while simultaneously collecting air pressure data using a high-frequency anemometer mounted on the end of the micro-orifice nozzle. In the control test, the comparative group used a single closed-loop negative pressure feedback control program, passively increasing the speed of the variable frequency negative pressure source only after detecting a decrease in suction at the micro-orifice nozzle end. The experimental group, however, used a time-advance coordinated control pulse excitation loop and variable frequency negative pressure source to perform feedforward aerodynamic compensation. The comparative indicators recorded by the data acquisition system showed that in the comparative group, within a 0.5-second time interval during a significant bend in the flexible duct, the negative pressure value at the end of the micro-orifice nozzle decreased from the standard working negative pressure of 2000 Pascals to 1200 Pascals, subsequently triggering an overshoot suction of 2800 Pascals during the duct flattening phase, resulting in a dynamic fluctuation of 40% in suction force. Within the same flexible duct bending cycle, the experimental group's collaborative drive module sent an action signal 45 milliseconds in advance, driving the piezoelectric ceramic micro-valve array to inject a high-pressure micro-airflow of 0.5 MPa tangentially. The feedforward pneumatic compensation action ensured that the negative pressure at the end of the micro-orifice nozzle remained stably maintained within the standard range of 1980 Pa to 2020 Pa. The experimental group's test data demonstrated that the dual physical intervention mechanism reduced the dynamic fluctuation of suction to less than 1%, objectively proving that the control device eliminated the suction instability caused by pure pneumatic hysteresis.

Claims

1. A negative pressure adsorption-type droplet recovery device for non-contact spraying, characterized in that, It includes equipment and control devices, with the control devices establishing a communication connection with the equipment. The equipment includes a liquid spraying unit, an annular negative pressure chamber, a flexible duct, a variable frequency negative pressure source, a position sensing component, a pulse excitation ring, and a broadband detector component. The annular negative pressure chamber is located around the liquid spraying unit and is connected to the beginning of the flexible duct. The end of the flexible duct is connected to the variable frequency negative pressure source. The pulse excitation ring is installed at the junction of the annular negative pressure chamber and the flexible duct. The position sensing component is located on the liquid spraying unit, and the broadband detector component is located on the inner wall of the flexible duct. The control device includes: Data acquisition module: used to obtain absolute spatial coordinates through position sensing components, extract motion trajectory parameters from absolute spatial coordinates, and convert motion trajectory parameters into spatial curvature vectors; Distortion calculation module: used to receive spatial curvature vector and broadband shock wave signal obtained by broadband detection component, use spatial curvature vector and broadband shock wave signal to deduce the location of flow field separation singularity, and output vortex blockage index according to fluid kinetic energy dissipation rate at the location of flow field separation singularity. Time Delay Prediction Module: Used to receive the vortex congestion index, calculate the transient resistance increment caused by the inward contraction deformation of the flexible duct under negative pressure, and map the transient resistance increment into a spatiotemporal multidimensional dynamic delay matrix. Waveform optimization module: used to receive the spatiotemporal multidimensional dynamic delay matrix, extract the fluid oscillation parameters in the spatiotemporal multidimensional dynamic delay matrix to generate an anti-phase compensation acoustic wave topology map, and output the jet excitation parameters and speed adjustment value according to the anti-phase compensation acoustic wave topology map; Collaborative drive module: Used to receive jet excitation parameters and speed adjustment values, send speed adjustment electrical signals to the frequency conversion negative pressure source, and synchronously send pulse excitation action electrical signals to the pulse excitation ring.

2. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 1, characterized in that, The data acquisition module includes a coordinate filtering unit and a base transformation unit; The coordinate filtering unit obtains the absolute spatial coordinates, uses the Kalman filter algorithm to filter out the mechanical vibration noise in the absolute spatial coordinates, and outputs the noise-reduced coordinate data. The base frame conversion unit receives denoised coordinate data, constructs a three-dimensional active frame using the denoised coordinate data, extracts the motion tangent vector, normal vector, and subnormal vector, calculates the first derivative of the motion tangent vector with respect to the arc length parameter to obtain the bending curvature, calculates the first derivative of the subnormal vector with respect to the arc length parameter to obtain the spatial deflection, and outputs the bending curvature and spatial deflection as a spatial curvature vector including bending curvature and spatial deflection to the distortion calculation module.

3. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 2, characterized in that, The distortion calculation module includes a frequency domain decoupling unit and a partial differential calculation unit; The frequency domain decoupling unit receives the broadband shock wave signal, performs a fast Fourier transform on the broadband shock wave signal, and extracts the characteristic frequency physical quantities. The partial differential equation unit synchronously receives the spatial curvature vector and the characteristic frequency physical quantity, maps the spatial curvature vector into a centrifugal inertial force term, introduces the centrifugal inertial force term into the momentum conservation boundary condition of the Navier-Stokes partial differential equation, solves the in-flow velocity partial derivative approaching the inner wall of the flexible duct, marks the spatial coordinate point where the in-flow velocity partial derivative is equal to zero as the flow field separation singularity position, extracts the amplitude of the characteristic frequency physical quantity at the flow field separation singularity position, and uses the amplitude to calculate the fluid kinetic energy dissipation rate at the flow field separation singularity position to establish the vortex congestion index.

4. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 3, characterized in that, The time delay prediction module includes a viscoelastic observation unit and a matrix mapping unit; The viscoelastic observation unit receives the vortex congestion index, retrieves the preset polymer material relaxation time constant, sets the polymer material relaxation time constant as the calculus order of the fractional calculus operator, inputs the vortex congestion index as the excitation source into the fractional calculus operator, and uses the fractional calculus operator to calculate the transient resistance increment caused by the inward contraction deformation of the flexible duct under the action of the vortex congestion index. The matrix mapping unit receives the transient resistance increment, divides the transient resistance increment into a discretized grid along the axial length of the flexible duct, extracts the spatial location parameters of the discretized grid, and extends and transforms the single time dimension delay into a spatiotemporal multidimensional dynamic delay matrix that includes the spatial location parameters.

5. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 4, characterized in that, The waveform optimization module includes a waveform inversion unit and a parameter allocation unit; The waveform inversion unit receives the spatiotemporal multidimensional dynamic delay matrix, extracts the fluid oscillation amplitude distribution map and phase angle distribution matrix from the spatiotemporal multidimensional dynamic delay matrix as fluid oscillation parameters, increases the values ​​of all elements in the phase angle distribution matrix by 180 degrees, and reconstructs the phase angle distribution matrix after increasing the values ​​by 180 degrees with the fluid oscillation amplitude distribution map to obtain the anti-phase compensation acoustic wave topology map. The parameter allocation unit receives the anti-phase compensation acoustic wave topology diagram, uses Hilbert transform to analyze the anti-phase compensation acoustic wave topology diagram to obtain the instantaneous analytical signal, extracts the amplitude envelope extreme value of the instantaneous analytical signal, calculates the amplitude envelope extreme value as the jet excitation parameter of the driving pulse excitation ring, and simultaneously converts the amplitude envelope extreme value into the speed adjustment value to offset the macroscopic resistance of the pipeline.

6. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 5, characterized in that, The jet excitation parameters calculated by the parameter allocation unit include the Strouhal locking frequency and the adaptive pulse width duty cycle; The parameter allocation unit uses the spatial curvature vector to calculate the pipe section contraction diameter of the flexible duct, sets the pipe section contraction diameter as the local airflow characteristic length, obtains the real-time air extraction volume of the variable frequency negative pressure source to calculate the mainstream time-averaged velocity, extracts the characteristic frequency physical quantity from the broadband shock wave signal, multiplies the characteristic frequency physical quantity by the local airflow characteristic length and divides it by the mainstream time-averaged velocity to obtain the Strouhal lock frequency.

7. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 6, characterized in that, The collaborative drive module includes a phase lead compensation unit and a frequency converter output unit; The phase advance compensation unit receives the jet excitation parameters and the spatiotemporal multidimensional dynamic delay matrix, extracts the time lag variable corresponding to the current spatial position in the spatiotemporal multidimensional dynamic delay matrix, sets the time lag variable as the time advance, and sends the pulse excitation action electrical signal including the jet excitation parameters to the pulse excitation ring in advance according to the time advance. The frequency converter output unit receives the speed adjustment value and the time advance, and sends the speed adjustment electrical signal to the frequency converter negative voltage source in advance according to the time advance.

8. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 7, characterized in that, The position sensing component includes a multi-axis grating ruler encoding component, which is installed on the side of the spraying unit to collect positioning pulse electrical signals and convert the positioning pulse electrical signals into absolute spatial coordinates; The broadband detection component includes a microelectromechanical high-frequency dynamic piezoelectric converter, which is flush-fitted into the bending stress stagnation point of the flexible duct inner wall, converting the collected flow field pressure pulsation into a continuous electrical waveform as a broadband shock wave signal.

9. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 8, characterized in that, The pulse excitation ring is surrounded by an array of piezoelectric ceramic micro valves, and the injection openings of the piezoelectric ceramic micro valve array are inclined along the direction parallel to the tangent of the inner wall of the flexible air duct. The piezoelectric ceramic micro-valve array receives pulse-excited action signals to open and close instantaneously. It utilizes the Coanda effect to inject high-pressure micro-airflow tangentially into the boundary layer of the flexible duct. The high-pressure micro-airflow cuts off the energy return zone of the flow field separation bubbles generated inside the flexible duct.

10. The negative pressure adsorption droplet recovery device for non-contact spraying according to claim 9, characterized in that, The variable frequency negative pressure source is equipped with a centrifugal exhaust component. The variable frequency negative pressure source receives the speed adjustment electrical signal and changes the stator power supply frequency of the centrifugal exhaust component to change the air volume. The changed air volume is used in conjunction with the high-pressure micro airflow injected by the piezoelectric ceramic micro valve array to reshape the boundary layer velocity distribution characteristics at the flow field separation singularity position, so that the inner wall of the flexible air duct is forced to change from a non-slip state to an approximately slip state.

Images