A method and system for polishing the surface of cosmetic glass bottles

Abstract

This invention discloses a method and system for polishing the surface of cosmetic glass bottles. The method includes: modifying a base material with gradient abrasive composites; using laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface to form a composite abrasive polishing part; assembling the composite abrasive polishing part onto the execution end of a six-axis linkage robotic arm, and performing zoned differentiated polishing based on the image recognition results of surface contaminant distribution density to obtain a preliminary polished glass bottle blank; synchronously blowing an inert gas flow with a microbubble generator onto the surface of the glass bottle blank, while simultaneously controlling the composite abrasive polishing part to perform micro-vibration to obtain a semi-finished glass bottle blank; placing the semi-finished glass bottle blank in a sealed cavity for spraying, and performing final polishing with pulsed ultrasonic vibration to obtain the finished cosmetic glass bottle. Using this invention, high-precision polishing of the curved surface of glass bottles can be achieved, improving surface cleanliness and gloss uniformity.

Description

Technical Field

[0001] This invention belongs to the field of cosmetic technology, specifically a method and system for polishing the surface of cosmetic glass bottles. Background Technology

[0002] Traditional glass bottle polishing processes for cosmetics are generally inefficient and lack precision, especially when dealing with complex curved surfaces. Poor fit between the polishing tool and the bottle often results in scratches or uneven gloss. Existing technologies mostly rely on manual polishing or fixed-grit grinding wheels, making it difficult to achieve dynamic pressure adjustment and precise removal of contaminants, easily leading to debris residue or secondary contamination. Furthermore, separating the routine cleaning steps from the polishing process increases the complexity of the process. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for polishing the surface of cosmetic glass bottles, so as to overcome the shortcomings of the prior art, and to achieve high-precision adaptive polishing of the curved surface of the glass bottle, thereby improving the surface cleanliness and gloss uniformity.

[0004] One embodiment of this application provides a method for polishing the surface of a cosmetic glass bottle, the method comprising:

[0005] The base material of sandpaper is modified with gradient abrasive composite, and three-dimensional cutting is performed according to the curved contour of the glass bottle surface using laser cutting technology to form a composite sandpaper polishing part that precisely fits the curved surface of the bottle.

[0006] The composite sandpaper polishing part is assembled on the execution end of a six-axis linkage robotic arm. The three-dimensional morphological data of the glass bottle surface is obtained by infrared scanning. Based on the morphological data, the rotation speed of the servo motor and the contact pressure of the polishing part are dynamically adjusted. Combined with the image recognition results of the surface contaminant distribution density, zoned differential polishing is carried out to obtain the glass bottle blank after preliminary polishing.

[0007] On the surface of the glass bottle blank after preliminary polishing, an inert gas flow with a microbubble generator is used for synchronous purging, while the composite sandpaper polishing parts are controlled to vibrate slightly. The residual polishing debris is removed through gas-solid coupling, resulting in a finely polished glass bottle semi-finished blank.

[0008] The finely polished glass bottle preform is placed in a sealed cavity and sprayed with a suspension of nano-sized silica flexible abrasive particles. Finally, it is polished by pulsed ultrasonic vibration. The residual micron-sized contaminants are removed through the synergistic effect of particle Brownian motion and ultrasonic cavitation, resulting in a cosmetic glass bottle with a clean surface and a preset gloss level.

[0009] Optionally, the process of modifying the base material of the sandpaper with gradient abrasive composite, and then using laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface to form a composite sandpaper polishing part that precisely fits the curved surface of the bottle, includes:

[0010] Natural sand-like material is placed in supercritical carbon dioxide fluid, and the fluid density is controlled by adjusting the pressure and temperature to achieve the regulation of the nanoscale pore structure of the sand-like material fibers, forming a porous carrier structure, which provides anchoring points for subsequent abrasive particle adhesion.

[0011] Electrophoretic deposition technology was used to sequentially deposit 800-mesh, 1500-mesh, and 3000-mesh zirconia abrasive grains on the surface of a porous carrier structure. The abrasive grain deposition rate was controlled by pulse current density to form a gradient abrasive grain layer, which constitutes a gradient abrasive composite substrate.

[0012] Using a femtosecond laser micro-nano processing system, contour tracking and cutting of a gradient abrasive composite substrate is performed based on the three-dimensional model data of a cosmetic glass bottle. At the same time, micro-serrated joints are generated at the cutting edge to complete the preparation of a three-dimensional pre-cut and polished part.

[0013] The three-dimensional pre-cut polishing part is immersed in a silane coupling agent solution for surface activation, and then hot-pressed with a shape memory polymer substrate. The shape memory effect of the substrate gives the polishing part an adaptive bonding ability, forming a composite sandpaper polishing part.

[0014] Optionally, the composite sandpaper polishing component is assembled onto the execution end of a six-axis linkage robotic arm. Three-dimensional morphological data of the glass bottle surface is acquired through infrared scanning. Based on this morphological data, the rotational speed of the servo motor and the contact pressure of the polishing component are dynamically adjusted. Combined with the image recognition results of the surface contaminant distribution density, differentiated polishing is performed in different zones to obtain a preliminary polished glass bottle blank, including:

[0015] An infrared thermal imager and a structured light scanner are integrated into the execution end of the robotic arm to perform multimodal scanning on the surface of the glass bottle. Subsurface defects are identified through thermal wave imaging technology, and the scan data is fused to generate a three-dimensional point cloud dataset containing morphology and defect information.

[0016] Based on a 3D point cloud dataset, a standard bottle model is matched using a dynamic time warping algorithm to calculate a local curvature deviation map. Combined with a deep learning semantic segmentation model, the types and distribution densities of surface contaminants are identified, and a differentiated polishing strategy map is constructed.

[0017] Using curvature deviation map and pollutant distribution density as input, the servo motor speed and contact pressure parameters are extrapolated in real time through fuzzy control rule base to form a dynamic control parameter set;

[0018] Based on the dynamic control parameter set, the six-axis robotic arm is driven to move along a preset trajectory and simultaneously perform zoned grinding operations. The dynamic error of the robotic arm is corrected through iterative learning control algorithms to complete the initial grinding and obtain the glass bottle blank.

[0019] Optionally, the surface of the glass bottle preform after preliminary polishing is simultaneously purged with an inert gas flow equipped with a microbubble generator, while the composite abrasive polishing component is controlled to vibrate slightly. This process removes residual polishing debris through gas-solid coupling, resulting in a finely polished semi-finished glass bottle preform. This includes:

[0020] Argon gas is introduced into a venturi tube and mixed with deionized water. Microbubbles of a specific diameter are generated at the gas-liquid interface through ultrasonic cavitation effect, forming a gas-liquid two-phase flow, which serves as a medium for debris removal.

[0021] The distribution of debris on the surface of the glass bottle is monitored in real time by a high-speed camera. Particle image velocimetry technology is used to calculate the trajectory of the debris. Based on the trajectory data, the injection angle and velocity of the gas-liquid two-phase flow are dynamically adjusted to form precise purging parameters.

[0022] Based on the purging parameters, the composite sandpaper grinding part is controlled to perform elliptical trajectory micro-vibration at a specific micro-vibration frequency. The amplitude and phase difference are optimized by vibration mode analysis to enhance the chip resonance detachment effect.

[0023] Ultraviolet radiation is introduced during the purging process to decompose organic residues through photocatalytic oxidation. Suspended particles are collected by an electrostatic adsorption device to complete the removal of residual debris and obtain a semi-finished glass bottle.

[0024] Optionally, the process involves placing the finely polished glass bottle semi-finished blank into a sealed cavity, spraying it with a suspension of nano-sized silica flexible abrasive particles, and performing final polishing using pulsed ultrasonic vibration. This process utilizes the synergistic effect of particle Brownian motion and ultrasonic cavitation to remove residual micron-sized contaminants, resulting in a finished cosmetic glass bottle with a clean surface and a preset gloss level.

[0025] Silica particles of a specific size are dispersed in deionized water, and an appropriate amount of polyvinylpyrrolidone is added as a dispersant. A stable suspension is prepared by ultrasonic homogenization and used as the grinding medium for final polishing.

[0026] A pressure-type atomizing nozzle is used to spray nano-grinding media onto the surface of a glass bottle in a fan-shaped spray pattern. The spray particle size distribution is monitored in real time by a laser particle size analyzer, and the spray pressure is dynamically adjusted to form a uniform liquid film.

[0027] Pulsed ultrasonic vibration with a specific pulse vibration frequency is applied to a glass bottle covered with a uniform liquid film. The generation and collapse cycle of cavitation bubbles are controlled by duty cycle modulation. Combined with fluid dynamics simulation, the transducer position is optimized to enhance the selective polishing effect.

[0028] The polishing effect is monitored in real time by an online gloss meter. The ultrasonic power and nanoparticle concentration are dynamically adjusted based on the monitoring data. When the gloss reaches the preset standard, the cleaning program is triggered. The polishing media is recycled through a circulating filtration system to obtain a finished cosmetic glass bottle with a clean surface and the preset gloss.

[0029] Another embodiment of this application provides a surface polishing system for cosmetic glass bottles, the system comprising:

[0030] The cutting module is used to modify the sandpaper base material with gradient abrasive composite. It uses laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface, forming a composite sandpaper polishing part that precisely fits the curved surface of the bottle.

[0031] The polishing module is used to assemble the composite sandpaper polishing parts onto the execution end of the six-axis linkage robotic arm. It acquires three-dimensional morphological data of the glass bottle surface through infrared scanning, dynamically adjusts the rotation speed of the servo motor and the contact pressure of the polishing parts based on the morphological data, and performs zoned differentiated polishing in combination with the image recognition results of the surface contaminant distribution density to obtain the glass bottle blank after preliminary polishing.

[0032] The purging module is used to simultaneously purge the surface of the glass bottle blank after preliminary polishing with an inert gas flow equipped with a microbubble generator, while controlling the composite sandpaper polishing parts to vibrate slightly. Through gas-solid coupling, residual polishing debris is removed to obtain a finely polished glass bottle semi-finished blank.

[0033] The spray module is used to place the finely polished glass bottle semi-finished blank into a sealed cavity, and spray it with a suspension of nano-sized silica flexible abrasive particles. Combined with pulsed ultrasonic vibration, the final polishing is carried out. The residual micron-sized contaminants are removed through the synergistic effect of particle Brownian motion and ultrasonic cavitation, resulting in a cosmetic glass bottle with a clean surface and a preset gloss level.

[0034] Another embodiment of this application provides a storage medium storing a computer program, wherein the computer program is configured to execute the method described in any of the preceding claims when running.

[0035] Another embodiment of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the method described in any of the preceding claims.

[0036] Compared with existing technologies, the present invention provides a method for polishing the surface of cosmetic glass bottles. This method involves modifying a base material with gradient abrasive mesh size, and then using laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface, forming a composite abrasive polishing part. The composite abrasive polishing part is then assembled onto the execution end of a six-axis linkage robotic arm, and differentiated polishing is performed in zones based on the image recognition results of surface contaminant distribution density, resulting in a preliminary polished glass bottle blank. On the surface of the glass bottle blank, an inert gas flow with a microbubble generator is used for synchronous blowing, while the composite abrasive polishing part is controlled to undergo micro-vibration, resulting in a semi-finished glass bottle blank. The semi-finished glass bottle blank is placed in a sealed cavity for spraying, and then subjected to final polishing with pulsed ultrasonic vibration to obtain the finished cosmetic glass bottle. This method achieves high-precision adaptive polishing of the curved surface of the glass bottle, improving surface cleanliness and gloss uniformity. Attached Figure Description

[0037] Figure 1 Hardware structure block diagram of a computer terminal for a method of polishing the surface of a cosmetic glass bottle provided in an embodiment of the present invention;

[0038] Figure 2 A schematic flowchart of a method for polishing the surface of a cosmetic glass bottle provided in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of a surface polishing system for cosmetic glass bottles provided in an embodiment of the present invention. Detailed Implementation

[0040] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0041] This invention first provides a method for polishing the surface of cosmetic glass bottles. This method can be applied to electronic devices, such as computer terminals, specifically ordinary computers.

[0042] The following detailed explanation uses a computer terminal as an example. Figure 1 This is a hardware structure block diagram of a computer terminal for a method of polishing the surface of a cosmetic glass bottle, provided in an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0043] Non-volatile storage media can store operating systems and computer programs. These computer programs include program instructions that, when executed, cause the processor to perform any method for polishing the surface of cosmetic glass bottles.

[0044] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0045] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to perform any method of polishing the surface of cosmetic glass bottles.

[0046] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0047] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0048] See Figure 2 The present invention provides a method for polishing the surface of a cosmetic glass bottle, which may include the following steps:

[0049] S201 involves modifying the base material of sandpaper with gradient abrasive composite, and using laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface to form a composite sandpaper polishing part that precisely fits the curved surface of the bottle.

[0050] Specifically, natural sand-like materials can be placed in supercritical carbon dioxide fluid, and the fluid density can be controlled by adjusting the pressure and temperature to achieve nanoscale channel structure regulation of sand-like fibers, forming a porous carrier structure, which provides anchoring points for subsequent abrasive particle adhesion;

[0051] Construction of Supercritical Fluid Processing System

[0052] Natural suede material (such as tanned leather substrate) is first cut into standard-sized blanks and loaded into a high-pressure reactor. This reactor is equipped with a precision temperature control system (temperature accuracy ±0.5 degrees Celsius) and a pressure sensor (range 0-50 MPa). Liquid carbon dioxide is injected into the reactor, and the temperature is raised to above the critical temperature of 31.1 degrees Celsius (typical operating temperature is 45 degrees Celsius) by a heater, while the pressure is raised to above the critical pressure of 7.38 MPa (typical operating pressure is 15 MPa) by a hydraulic pump. In this state, the carbon dioxide enters the supercritical state (SC-ST), and its density can be continuously adjusted within the range of 300-900 kg / m³. By adjusting the pressure and temperature in real time (e.g., increasing the pressure in steps from 12 MPa to 18 MPa, and the temperature slowly increasing from 40 degrees Celsius to 50 degrees Celsius), the supercritical carbon dioxide fluid density is controlled at a target value (e.g., 600 kg / m³), which directly affects the fluid's ability to swell and permeate the suede fibers.

[0053] Nanopore regulation mechanism

[0054] Under the osmotic effect of supercritical carbon dioxide, the collagen molecular chains in the cortical fibers undergo directional relaxation. The fluid's high diffusivity (diffusion coefficient approximately 10) contributes to this process. -7 The fluid penetrates deep into the fiber microstructure at a pressure of (square meters per second), preferentially dissolving and removing lipid impurities and low-molecular-weight non-collagenous proteins. By maintaining a high-pressure state (e.g., 15 MPa) for 30 minutes, the natural pores inside the fiber expand under the action of fluid tension, forming uniformly distributed nano-channels (NC) with a pore size range of 50-200 nanometers. The pore density and size are precisely controlled by the fluid density: at a density of 600 kg / m³, dense channels with an average pore size of 80 nanometers are generated; when the density increases to 800 kg / m³, the pore size expands to 150 nanometers, but the density decreases. This process simultaneously completes the degreasing and purification of the material, ultimately forming a porous carrier structure with an open three-dimensional network (Porosity > 65%).

[0055] Anchor positioning point function enhancement

[0056] After being removed from the reactor, the porous support structure was subjected to gradient depressurization (depressurization rate 0.5 MPa per minute) to prevent pore collapse. After verifying the pore morphology using scanning electron microscopy (SEM), it was immersed in a 0.5 mol / L sodium hydroxide solution for surface hydroxylation treatment, generating a large number of active hydroxyl groups (-OH) on the inner walls of the pores. These hydroxyl groups act as anchoring sites (AS), significantly enhancing the chemical bonding strength of subsequent abrasive particles. The pore depth was controlled by the treatment time: a 30-minute treatment yielded an active pore layer approximately 5 micrometers deep, while a 60-minute treatment expanded it to 12 micrometers, ensuring effective mechanical locking after abrasive particle embedding.

[0057] Electrophoretic deposition technology was used to sequentially deposit 800-mesh, 1500-mesh, and 3000-mesh zirconia abrasive grains on the surface of a porous carrier structure. The abrasive grain deposition rate was controlled by pulse current density to form a gradient abrasive grain layer, which constitutes a gradient abrasive composite substrate.

[0058] Electrophoretic deposition system configuration

[0059] A porous carrier structure was fixed to a cathode electrode plate and placed in an electrophoresis tank. The tank was filled with a suspension composed of zirconia abrasive grains (particle size: 800 mesh approximately 21 micrometers, 1500 mesh approximately 10 micrometers, 3000 mesh approximately 5 micrometers), isopropanol (dispersion medium), and cerium nitrate (conductive additive). The solid content of the suspension was controlled at 20% by weight, and the pH was adjusted to 8.5 with ammonia to ensure that the abrasive grain surface carried a positive charge (Zeta potential +35 mV). The deposition process was carried out in three stages, with the corresponding mesh size suspension replaced in each stage.

[0060] Pulse current deposition control

[0061] The first stage involves depositing 800-mesh abrasive grains: a pulsed current (PC) is applied with parameters set to a peak current density of 15 mA / cm², a pulse width of 50 ms, and an interval of 100 ms. The high current density promotes rapid migration of large-diameter abrasive grains, forming a coarse abrasive layer approximately 30 micrometers thick within 50 seconds. The second stage involves switching to a 1500-mesh suspension: the pulse parameters are adjusted to a peak current density of 8 mA / cm², a pulse width of 80 ms, and an interval of 50 ms. Under this moderate current density, the abrasive grains stack in an orderly manner, forming a 20-micrometer thick intermediate transition layer. The third stage uses a 3000-mesh suspension: a low peak current density of 5 mA / cm², a pulse width of 100 ms, and an interval of 30 ms are employed. By extending the pulse duration, dense filling of fine abrasive grains is achieved, generating a fine surface layer 10 micrometers thick.

[0062] Gradient structure solidification

[0063] After each deposition stage, the operation is paused, and unbonded abrasive grains are shaken off using a micro-vibration table (50 μm amplitude, 100 Hz frequency). After the three layers are deposited, the entire substrate is immersed in a silica sol binder and heat-treated at 120°C for 1 hour. The silica sol dehydrates to form a siloxane network, which covalently bonds the abrasive grains to the carrier fibers, ultimately forming a graded abrasive composite substrate (GACS). The abrasive grain size transitions continuously from 5 μm to 10 μm to 21 μm from the surface to the bottom layer, and the interlayer bonding strength is tested to reach 15 MPa.

[0064] Using a femtosecond laser micro-nano processing system, contour tracking and cutting of a gradient abrasive composite substrate is performed based on the three-dimensional model data of a cosmetic glass bottle. At the same time, micro-serrated joints are generated at the cutting edge to complete the preparation of a three-dimensional pre-cut and polished part.

[0065] Laser processing system calibration

[0066] The CAD surface model of the cosmetic glass bottle was imported into a femtosecond laser system (FLS) equipped with a 1030 nm wavelength laser source (pulse width 350 femtoseconds, repetition rate 100 kHz). First, coordinate calibration was performed: a fluorescent positioning layer was sprayed onto the surface of a gradient abrasive composite substrate, and images of the substrate were acquired using a CCD vision system and spatially registered with the glass bottle model (positioning accuracy ±5 μm). The processing area was divided according to the surface curvature distribution; a fine path spacing of 0.05 mm was used in high curvature areas (curvature radius < 20 mm), while the path spacing was extended to 0.2 mm in low curvature areas (curvature radius > 50 mm).

[0067] Contour tracking and micro-sawing

[0068] The laser focus is consistently maintained 10 micrometers below the substrate surface via a dynamic focusing module. During cutting along the 3D model contour, a Contour Tracking Algorithm (CTA) is used to adjust the laser scanning speed in real time: 800 mm / s for straight sections and reduced to 400 mm / s for circular sections. Edge modification is performed simultaneously with the cutting: 1 MHz oscillating scans are superimposed on both sides of the cutting path, increasing the laser energy density to 8 joules per square centimeter, causing selective ablation of the material and forming a micro-serrated joint (MSJ) with a depth of approximately 50 micrometers and a pitch of 20 micrometers. This structure increases the contact area during subsequent lamination.

[0069] 3D pre-cut parts forming

[0070] After cutting, the substrate is divided into multiple curved surface units (such as independent units for the shoulder, belly, and bottom) that perfectly match the bottle's surface. The cut parts are separated using a negative pressure adsorption transfer device, and a white light interferometer is used to inspect the edge serration morphology (tooth height tolerance ±2 micrometers) and contour fit (deviation <25 micrometers). The final result is a 3D pre-cut abrading component (3D-PAC) that retains the gradient abrasive structure, with a maximum deviation of no more than 0.5% from the target glass bottle's surface curvature.

[0071] The three-dimensional pre-cut polishing part is immersed in a silane coupling agent solution for surface activation, and then hot-pressed with a shape memory polymer substrate. The shape memory effect of the substrate gives the polishing part an adaptive bonding ability, forming a composite sandpaper polishing part.

[0072] Surface activation treatment

[0073] Prepare a 3% (volume ratio) ethanol solution of KH-550 silane coupling agent (aminopropyltriethoxysilane). Immerse the three-dimensional pre-cut abrasive part in the solution and treat it under ultrasonic assistance (300 W, 40 kHz) for 20 minutes to allow the coupling agent molecules to penetrate into the gaps between the abrasive grains. After removal, pre-curing at 80°C for 15 minutes forms a chemically activated layer. The activated layer contains two functional groups: siloxane (-Si-O-) which is bonded to the abrasive / carrier, and amino (-NH2) groups which provide reaction sites facing outwards.

[0074] Shape memory polymer composite

[0075] A 0.5 mm thick polyurethane-based shape memory polymer substrate (SMPS) with a glass transition temperature (Tg) set to 55 degrees Celsius was cut. The activated polished part was placed on the SMPS and fed into a hot press. First stage: The temperature was raised to 70 degrees Celsius (15 degrees Celsius above Tg), and a pressure of 0.3 MPa was applied and held for 5 minutes to soften the SMPS and embed it into the micro-serrated structure of the polished part. Second stage: The temperature was lowered to 40 degrees Celsius (15 degrees Celsius below Tg), and the pressure was increased to 0.8 MPa and held for 10 minutes to achieve mechanical interlocking. At this point, the SMPS entered a temporary shape state (TSS).

[0076] Adaptive fit capability gives

[0077] The composite material undergoes shape programming: it is reheated to 70 degrees Celsius and deformed into a planar state under the constraint of an outer mold. After cooling and fixing, a pre-set strain (PS) is formed. When applied to a glass bottle at 40 degrees Celsius, the SMPS partially recovers due to its temperature approaching Tg, generating a progressive bonding pressure of approximately 0.5 Newtons per square centimeter. This shape memory effect (SME) allows the polishing part to automatically adapt to changes in the curvature of the bottle (adapting to curvature radii ranging from 15 to 200 mm), ultimately forming a composite leather abrasive tool (CLAT) that combines gradient grinding capabilities with surface self-adaptation.

[0078] S202, the composite sandpaper grinding part is assembled on the execution end of the six-axis linkage robotic arm. The three-dimensional morphological data of the glass bottle surface is obtained by infrared scanning. Based on the morphological data, the rotation speed of the servo motor and the contact pressure of the grinding part are dynamically adjusted. Combined with the image recognition results of the surface contaminant distribution density, the partitioned differential grinding is carried out to obtain the glass bottle blank after preliminary grinding.

[0079] Specifically, an infrared thermal imager and a structured light scanner can be integrated into the execution end of the robotic arm to perform multimodal scanning of the glass bottle surface, identify subsurface defects through thermal wave imaging technology, and fuse the scan data to generate a three-dimensional point cloud dataset containing morphology and defect information.

[0080] The end effector of the six-axis robotic arm simultaneously assembles a composite sanding surface polishing component, an infrared thermal imager (ITI), and a structured light scanner (SLS) via a customized fixture. The ITI employs a mid-wave infrared detector (wavelength range 3-5 micrometers), its core being a mercury cadmium telluride (HgCdTe) focal plane array with a pixel size of 15 micrometers and a thermal sensitivity of 0.03 degrees Celsius. The SLS projects a combination of Gray code and phase-shifted fringe patterns using a blue LED light source (wavelength 450 nanometers), coupled with a 5-megapixel industrial camera to capture deformation fringes. Upon initiation of the scan, the robotic arm moves along a preset path around the glass bottle, and the ITI acquires the surface temperature field distribution at a rate of 30 frames per second. When subsurface cracks or bubbles exist in the glass bottle, its thermal conductivity differs from normal areas, manifesting as localized temperature differences in the ITI image (e.g., a temperature 0.5 degrees Celsius lower at the crack). The SLS simultaneously projects the grating pattern and acquires the deformation fringes, reconstructing the surface's geometric contours with millimeter-level precision using a phase unwrapping algorithm.

[0081] Thermal Wave Imaging (TWI) relies on pulsed thermal excitation: before scanning, a robotic arm controls an array of halogen lamps (1000 watts) integrated into a fixture to apply a short 0.5-second thermal pulse to the surface of the glass bottle. ITI then records the temperature decay curve of the surface cooling process (sampling interval 0.1 seconds). Subsurface defects alter the heat diffusion rate; for example, a microcrack with a depth of 0.2 mm can accelerate the cooling rate of the corresponding area by 15%. By extracting the second derivative features of the temperature decay curve and combining them with a pre-trained defect recognition model (based on a convolutional neural network CNN), the thermal image is transformed into a subsurface defect probability map (DPM). Simultaneously, the 3D point cloud data from SLS (accuracy ±0.05 mm) is spatially registered with the DPM: using the robotic arm's base coordinate system as a reference, point cloud fusion is achieved through an iterative closest point (ICP) algorithm, ultimately generating an enhanced 3D point cloud dataset containing 500,000 to 1 million data points. Each data point contains spatial coordinates (X,Y,Z), normal vector (Nx,Ny,Nz), local curvature (C), and defect marker (D value 0-1 indicates defect probability).

[0082] The data fusion process employs a multi-resolution pyramid architecture: the original SLS point cloud is first downsampled to 1 mm resolution using a voxel mesh as the base layer, while the ITI defect probability map is upsampled to the same resolution and mapped onto the base layer point cloud. Subsequently, the high-resolution (0.1 mm) SLS detail layer is superimposed with the defect information layer using a Laplacian pyramid fusion algorithm, preserving geometric details while embedding defect attributes. The fused point cloud dataset is stored in an octree structure, supporting real-time spatial queries. For example, when the normal vector of a point cloud region abruptly changes by more than 15 degrees and the D value is greater than 0.8, the system automatically labels it as a "high-probability subsurface defect area," which directly affects the formulation of subsequent polishing strategies. The entire process is completed within a single orbital motion of the robotic arm (approximately 12 seconds), providing a data foundation for real-time control.

[0083] Based on a 3D point cloud dataset, a standard bottle model is matched using a dynamic time warping algorithm to calculate a local curvature deviation map. Combined with a deep learning semantic segmentation model, the types and distribution densities of surface contaminants are identified, and a differentiated polishing strategy map is constructed.

[0084] The Standard Bottle Model (SBM) is stored in the database as a NURBS surface with a control point density of 5 points per square centimeter. The matching process first involves initial alignment of the enhanced point cloud dataset with the SBM: seven feature points, including the bottle neck and bottom, are selected, and the initial transformation matrix is ​​calculated using least-squares fitting. Then, Dynamic Time Warping (DTW) is used to handle non-rigid deformation: the point cloud is divided into 120 generatrices along the meridian direction, each containing 200 points. The DTW algorithm is applied to each generatrice, and by constructing a cumulative cost matrix (CostMatrix Size 200×200), the optimal nonlinear matching path between the point cloud and the corresponding generatrice of the SBM is found. This algorithm allows for local time axis scaling, effectively compensating for glass bottle manufacturing tolerances (such as a diameter deviation of ±0.3 mm). After matching, the positional deviation vectors (ΔX, ΔY, ΔZ) of each point are output, and then the local curvature deviation (CD) is calculated: 3×3 neighborhood points are extracted from the point cloud, and the principal curvatures K1 and K2 are obtained through least-squares surface fitting. The deviation value δK (unit: meters) is obtained by comparing it with the theoretical curvature of the SBM. -1 ).

[0085] Surface contaminant identification employs an end-to-end deep learning framework: First, the ResNet-50 backbone network is initialized through transfer learning, with input data consisting of RGB texture maps (2048×2048 resolution) acquired via SLS. The training dataset contains five typical contaminant classes: oil stains (Class 1), fingerprints (Class 2), adhesive residue (Class 3), dust (Class 4), and water stains (Class 5). The network output layer connects to a Conditional Random Field (CRF) to optimize spatial continuity, ultimately generating a pixel-level semantic segmentation map (SSM). Contaminant distribution density (CDD) is calculated based on sliding window statistics: the surface is divided into a 2 mm × 2 mm grid, and the pixel proportion of each contaminant class within each grid is statistically analyzed. For example, when the proportion of Class 2 pixels in a certain grid exceeds 60%, it is marked as a "high-density fingerprint contamination area."

[0086] The construction of the Differentiated Polishing Strategy Map (DPSM) is a three-dimensional spatial decision-making process: using point cloud data as the carrier, each point is assigned a four-dimensional decision vector (R, P, T, M). Wherein:

[0087] R-value (speed coefficient): set based on curvature deviation δK, when |δK|>0.5 meters. -1 In the high curvature variation zone, R=1.2 (corresponding to the servo motor's base speed of 8000 rpm × 1.2), and in the flat zone, R=0.8;

[0088] P-value (contact pressure): dynamically adjusted according to the type of contaminant; oil stains (Class 1) require high pressure (P=1.5 N), while dust (Class 4) requires low pressure (P=0.7 N).

[0089] T-value (polishing time): determined by CDD; for areas with density > 40%, T = twice the baseline time.

[0090] M value (motion mode): Subsurface defect area (D>0.8) triggers avoidance mode (M=0).

[0091] The map is spatially continuous through three-dimensional kriging interpolation, ultimately generating a digital strategy mesh with a resolution of 0.5 mm, which serves as the direct basis for the robotic arm's execution.

[0092] Using curvature deviation map and pollutant distribution density as input, the servo motor speed and contact pressure parameters are extrapolated in real time through fuzzy control rule base to form a dynamic control parameter set;

[0093] The Fuzzy Rule Base (FRB) employs a Mamdani-type inference architecture, containing two input variables (curvature deviation δK, contaminant density CDD) and two output variables (speed correction coefficient ΔR, pressure correction coefficient ΔP). Input variable fuzzification settings:

[0094] δK (unit: meter) -1 ): Divided into 5 fuzzy sets {negative large (NB), negative small (NS), zero (ZE), positive small (PS), positive large (PB)}, with a universe of discourse [-1.0, 1.0], and triangular membership functions;

[0095] CDD (unit: percentage): divided into 3 fuzzy sets {low (L), medium (M), high (H)}, universe of discourse [0, 100%], trapezoidal membership function;

[0096] The universes of discourse for the output variables ΔR and ΔP are both [-0.5, 0.5], divided into 7 fuzzy sets, and the membership functions are Gaussian.

[0097] The rule base contains 15 expert-based rules, for example:

[0098] Rule 7: If δK is PB (positive large curvature) and CDD is L (low pollution) → then ΔR = PS (positive small growth rate), ΔP = NS (negative small decompression rate);

[0099] Rule 12: If δK is ZE (zero curvature) and CDD is H (high pollution) → then ΔR = ZE (zero speed adjustment), ΔP = PB (positive pressure).

[0100] During real-time inference, parallel computation is performed on 2000 grid points per second. Using data from a specific point (δK = 0.6 meters)... -1 Taking CDD=75% as an example: δK activates PS (membership 0.7) and PB (0.3), and CDD activates H (1.0). Rule 10 (PS & H → ΔR=ZE, ΔP=PM) and rule 11 (PB & H → ΔR=NS, ΔP=PB) are triggered. Through weighted average defuzzification, the final output is ΔR=-0.12, ΔP=0.41. Combined with the baseline parameters (R0=8000 rpm, P0=1.0 N), the real-time command is obtained: speed 7040 rpm, pressure 1.41 N.

[0101] The Dynamic Control Parameter Set (DCPS) is stored as a spatiotemporal four-dimensional matrix (X, Y, Z, Time). The spatial dimension is aligned with the policy graph grid, and the time dimension is sampled at a frequency of 10 Hz. Parameter transmission uses the real-time Ethernet protocol (EtherCAT) with a cycle time of 1 millisecond. A key innovation lies in the introduction of an environmental compensation module: a temperature sensor monitors the workshop ambient temperature (T_env), and when T_env > 30℃, compensation rules are automatically injected (e.g., if increased temperature causes abrasive softening, pressure needs to be increased by 10%). After each bottle polishing is completed, the system automatically records the parameter optimization effect and updates the fuzzy rule weights through reinforcement learning. For example, if a high-pressure strategy targeting Class 3 contamination causes scratches, the confidence of the corresponding rule is reduced by 15%.

[0102] Based on the dynamic control parameter set, the six-axis robotic arm is driven to move along a preset trajectory and simultaneously perform zoned grinding operations. The dynamic error of the robotic arm is corrected through iterative learning control algorithms to complete the initial grinding and obtain the glass bottle blank.

[0103] The motion trajectory planning of the six-axis robotic arm (5 kg load, repeatability ±0.02 mm) uses B-spline curve interpolation. The standard trajectory library contains 300 basic paths for 12 bottle types, such as "shoulder spiral (Pitch=2 mm)" and "bottle body meridian reciprocation (amplitude ±15 degrees)". During execution, the spatial coordinates of the dynamic control parameter set (DCPS) are matched with the trajectory points in real time: when the end effector of the robotic arm reaches a certain spatial grid (e.g., coordinates X120, Y50, Z200), the (R, P, T, M) parameters of that position are extracted from the DCPS and converted into execution commands. Contact pressure control is achieved through a closed loop using a six-dimensional force sensor (range ±200 N, accuracy 0.1 N): when the deviation between the measured pressure and the target value exceeds 0.3 N, the impedance controller adjusts the Z-axis offset of the robotic arm within 50 milliseconds (adjustment range ±0.5 mm).

[0104] The core of zoned differentiated polishing lies in the allocation of time and space resources: dividing the bottle surface into 20 working zones (such as the bottom ring zone, the main body zone, and the neck transition zone). High curvature zone (|δK|>0.4 meters) -1 The system employs a "high-frequency, small-step" strategy (linear speed 40 mm / s, step size 0.8 mm) for high-contamination areas and a "low-frequency, large-step" strategy (linear speed 80 mm / s, step size 2 mm) for flat areas. For highly contaminated areas (CDD > 60%), a "three-stage grinding" process is implemented: the first stage is high-pressure coarse grinding (P = 1.5 N), the second stage is medium-pressure fine grinding (P = 1.0 N), and the third stage is low-pressure polishing (P = 0.5 N). The duration of each stage is determined by the T value. The robotic arm avoids subsurface defect areas (M = 0) in real time during execution: when a defect marker is detected, an avoidance path with a radius of 5 mm is automatically generated, simultaneously triggering compensation grinding in adjacent areas.

[0105] Iterative Learning Control (ILC) is used to correct system errors: Using the grinding trajectory of the first bottle as a reference, the positional error e_k(t) of each trajectory point is recorded (t is a time variable). After grinding, the actual amount removed is measured using a laser tracker (accuracy ±0.005 mm), and compared with the target amount removed to obtain the material removal error ε_k(t). According to the error model: u_{k+1}(t) = u_k(t) + L * e_k(t) + Q * dε_k / dt, where L=0.3 is the learning gain and Q=0.15 is the differential gain.

[0106] The control input u_{k+1}(t) for the next bottle is updated. For example, if the removal amount in the bottleneck area is less than 5 micrometers, an additional velocity pulse (increment of 8%) is injected into the trajectory point in that area in the next cycle. After 3-5 iterations, the position error is reduced to ±0.01 mm, and the material removal uniformity is improved to 95%. The final output surface roughness Ra of the glass bottle preform is reduced from the initial 3.2 micrometers to 0.8 micrometers, laying the foundation for subsequent fine polishing.

[0107] S203, on the surface of the glass bottle blank after preliminary polishing, an inert gas flow with a microbubble generator is used for synchronous purging, while the composite sandpaper polishing parts are controlled to vibrate slightly. The residual polishing debris is removed through gas-solid coupling, and a semi-finished glass bottle blank after fine polishing is obtained.

[0108] Specifically, argon gas can be introduced into a venturi tube and mixed with deionized water. Through ultrasonic cavitation, microbubbles of a specific diameter are generated at the gas-liquid interface to form a gas-liquid two-phase flow, which serves as a medium for debris removal.

[0109] After initial polishing, the surface of the rough glass bottle blank is covered with glass fragments of varying sizes, abrasive residue, and trace amounts of oil. The cleaning process first requires the preparation of an efficient cleaning medium—a gas-liquid two-phase flow. This system uses high-purity argon (Ar) as an inert carrier, with a purity of 99.999% (i.e., 5N grade) to prevent oxygen from participating and causing oxidation or watermarks on the glass surface. Argon is supplied from a high-pressure cylinder, and after being stabilized at 0.5 MPa by a precision pressure reducing valve, it is introduced into the inlet of a Venturi tube. A Venturi tube is a converging-diffraction pipe with a variable cross-section and an annular slit at its throat. Deionized water (DIW, resistivity ≥18 megohm·cm) is injected into the Venturi tube inlet by a centrifugal pump at a flow rate of 3 liters / minute (L / min). When a high-speed argon gas flow (approximately 15 m / s) passes through the throat, a local negative pressure (approximately -0.2 MPa) is generated according to Bernoulli's principle, forcefully drawing in deionized water and atomizing it into micron-sized droplets. The gas-liquid mixture undergoes thorough turbulent mixing in the diffusion section, forming a preliminary gas-liquid mixed flow.

[0110] The initially mixed gas-liquid flow then enters the ultrasonic cavitation reaction chamber. This chamber is made of titanium alloy with a corrosion-resistant ceramic coating on its inner walls, and its bottom integrates a piezoelectric ceramic ultrasonic transducer array (UTA). The transducers operate at a frequency of 40 kHz, driven by a high-frequency power supply to generate ultrasonic waves with a peak power of 800 watts (W). As the sound waves propagate through the fluid, they induce an ultrasonic cavitation effect at the gas-liquid interface: the negative pressure phase of the sound waves causes the local pressure of the fluid to fall below the saturated vapor pressure, forming tiny cavities (cavitation nuclei); the positive pressure phase causes the cavities to collapse rapidly, instantaneously generating local high pressures exceeding 1000 atmospheres and local high temperatures of 5000 Kelvin (K). These extreme physical conditions cause the argon gas to be efficiently sheared and broken up, combining with water molecules to form microbubbles with highly uniform diameters. By adjusting the ultrasonic power (adjustable from 300-800 W) and the cavitation chamber pressure (0.1-0.3 MPa), the bubble diameter can be precisely controlled within a specific range of 50±5 micrometers (μm). Bubbles of this size have optimal surface adsorption capacity and rising stability, effectively encapsulating and carrying debris.

[0111] The resulting gas-liquid two-phase flow (GLTPF) consists of 80% by volume microbubbles and 20% deionized water as a continuous phase. This fluid is delivered through a pressure-resistant pipe to a porous nozzle mounted at the end of a robotic arm. The nozzle has an orifice diameter of 200 micrometers and a pore density of 50 pores per square centimeter, ensuring that the two-phase flow uniformly covers the glass bottle surface. The deionized water acts as a wettant and buffer, preventing high-hardness glass fragments from scratching the surface; the argon microbubbles, with their huge specific surface area (approximately 12,000 square centimeters / ml), adsorb nano- to micron-sized debris and lift it off the surface by buoyancy. The entire medium generation process is monitored by a PLC (Programmable Logic Controller), with key parameters such as argon flow rate (5 liters / minute under standard conditions), water flow rate, ultrasonic power, and outlet bubble size (monitored online via a laser diffractometer) all subject to closed-loop feedback control.

[0112] The distribution of debris on the surface of the glass bottle is monitored in real time by a high-speed camera. Particle image velocimetry technology is used to calculate the trajectory of the debris. Based on the trajectory data, the injection angle and velocity of the gas-liquid two-phase flow are dynamically adjusted to form precise purging parameters.

[0113] To achieve targeted removal, a high-speed camera (HSC) is mounted above the polishing station. This camera uses a CMOS sensor with a resolution of 1920×1080 pixels and a frame rate of up to 2000 frames per second (fps), and is equipped with a ring-shaped LED cold light source (5600K color temperature) to provide shadowless illumination. The camera is tilted at a 60-degree angle towards the glass bottle surface to capture the surface microstructure in real time. The captured images are transmitted to the Image Processing Unit (IPU) for preprocessing, including Gaussian filtering for noise reduction, adaptive threshold segmentation to extract debris features, and morphological opening operations to remove artifacts. In the processed binary image, each connected component represents a debris particle, and its coordinates are recorded using a centroid localization algorithm. The system completes a full-area scan every 0.1 seconds, generating a real-time updated Debris Distribution Heatmap (DH), marking the location, size (equivalent diameter), and aggregation density (number of particles per unit area) of the debris.

[0114] To predict debris detachment behavior, the system employs Particle ImageVelocimetry (PIV). This technique is based on an optical flow algorithm: the instantaneous velocity vector v (in millimeters per second) is obtained by dividing the displacement vector Δd of the same debris in two consecutive frames (time interval Δt = 0.5 milliseconds) by Δt. By calculating the velocity field of tens of thousands of debris particles across the entire field, a Debris Trajectory Model (DTM) is constructed. This model can predict three key types of motion: 1) the downward trend caused by gravity; 2) radial migration caused by airflow drag; and 3) random deflection caused by inter-particle collisions. For example, a 10-micrometer diameter glass shard is expected to move 8 millimeters in 0.3 seconds in an airflow of 0.2 m / s, while a 50-micrometer debris, due to its greater inertia, will only move 2 millimeters. The model output is a dynamic Trajectory Probability Cloud (TPC), identifying the areas where debris may be distributed within the next second.

[0115] Based on the debris distribution heat map (DDH) and trajectory probability cloud map (TPC), the Dynamic Blowing Control System (DBCS) generates precise purging parameters (PBP). The system employs a fuzzy rule base for decision-making: 1) For high-density debris areas (>50 particles / mm²), the nozzle is instructed to tilt at a 45-degree angle and increase the flow rate to 0.8 m / s for impact purging; 2) For low-density areas (<10 particles / mm²), the nozzle is positioned vertically and the flow rate is reduced to 0.3 m / s to avoid over-wetting; 3) Based on trajectory prediction, a pre-interception mode is initiated for debris about to fall into the bottle opening—the nozzle is turned upstream 0.5 seconds in advance to intercept it. All parameters are adjusted via a servo motor to change the nozzle pitch angle (continuously adjustable from 0-90 degrees) and a proportional valve to regulate the gas-liquid flow rate (accuracy ±1%). The instructions are refreshed every 0.2 seconds to ensure that the purging strategy always matches the real-time operating conditions.

[0116] Based on the purging parameters, the composite sandpaper grinding part is controlled to perform elliptical trajectory micro-vibration at a specific micro-vibration frequency. The amplitude and phase difference are optimized by vibration mode analysis to enhance the chip resonance detachment effect.

[0117] During the two-phase gas-liquid purging process, the Composite Sandpaper Polishing Tool (CSPT) is equipped with micro-vibration functionality for synergistic effect. This polishing tool is mounted on the end flange of a six-axis robotic arm via a quick-change fixture, and its back integrates a piezoelectric actuator array (PAA). The actuator consists of four lead zirconate titanate (PZT-8) piezoelectric ceramic plates arranged in a cross shape, each capable of independent extension and retraction. When a sinusoidal voltage signal with a 90-degree phase difference is applied to two ceramic plates along the X-axis, they will synthesize an elliptical trajectory motion (ETM) in the plane. The vibration parameters are dynamically triggered by the purging parameters (PBP): high purging velocity regions (>0.6 m / s) correspond to high-frequency vibration (800 Hz), while low-velocity regions correspond to low-frequency vibration (200 Hz), with vibration energy and fluid kinetic energy complementing each other.

[0118] The key parameters of the elliptical trajectory—major axis amplitude A (range 1-10 μm), minor axis amplitude B (0.5-5 μm), and phase difference Φ (0-180 degrees)—are optimized using Vibration Modal Analysis (VMA). The specific process is as follows: 1) A miniature accelerometer (range ±50g, resolution 0.001g) is attached to the surface of the workpiece to collect vibration signals in real time; 2) The spectrum is decomposed using Fast Fourier Transform (FFT) to identify the mode shapes at the dominant frequencies; 3) The optimal parameter combination is calculated based on a chip removal efficiency model: for metal abrasives with strong adhesion (adhesion approximately 100 μN), a large amplitude (A=8 μm) and high ellipticity (A / B=2.5) "flat ellipse" is used to induce shearing and peeling; for lightweight glass powder (adhesion 10 μN), a small amplitude (A=2 μm) and high frequency (600 Hz) "slender ellipse" is used to induce resonant peeling. The optimized parameter set is called the Vibration Enhancement Factor (VEF).

[0119] Micro-vibration enhances debris removal through two mechanisms: 1) Resonance effect: When the vibration frequency approaches the natural frequency of the debris-matrix system (approximately 300-1000 Hz), it induces debris resonance, allowing it to overcome van der Waals forces. For example, a 20-micrometer-diameter glass shard can accelerate to 200g under 500 Hz vibration, generating a removal force more than 100 times its own weight. 2) Interfacial shear: The elliptical trajectory generates a tangential motion component on the surface of the workpiece, applying a continuous shear force to the debris. Combined with the buoyancy effect of the gas-liquid two-phase flow, this makes it easier for debris to enter the fluid and be carried away. Experimental results show that micro-vibration can increase the debris removal rate by 55% while reducing the amount of purging fluid by 40%. During vibration, an accelerometer continuously monitors the actual motion trajectory, and a PID controller (proportional-integral-derivative controller) dynamically compensates for the hysteresis effect of the piezoelectric ceramic, ensuring vibration accuracy better than ±0.2 micrometers.

[0120] Ultraviolet radiation is introduced during the purging process to decompose organic residues through photocatalytic oxidation. Suspended particles are collected by an electrostatic adsorption device to complete the removal of residual debris and obtain a semi-finished glass bottle.

[0121] To address potential organic contaminants (such as lubricating oil and fingerprint grease) that may remain during the polishing process, the system integrates an Ultraviolet Irradiation Module (UVIM) at the purging station. This module uses a low-pressure mercury lamp as the light source, with a dominant wavelength of 254 nanometers (nm) and a radiation intensity of 30 milliwatts per square centimeter (mW / cm²). 2Ultraviolet light, focused by a quartz lens, covers the entire surface of the glass bottle at a 45-degree angle. Under ultraviolet photon excitation, nano-titanium dioxide photocatalysts (20 nm particle size, 0.1% concentration), pre-sprayed onto the bottle surface, generate electron-hole pairs. Holes (h + It possesses strong oxidizing properties, directly decomposing the C-C bonds of organic matter; simultaneously, it reacts with water molecules in the air to generate hydroxyl radicals (·OH), with an oxidation potential reaching 2.8 volts (V), which can non-selectively degrade pollutants such as grease and resin. A typical reaction time of 3-5 seconds is sufficient to break the carbon chains of oil stains into CO2 and H2O.

[0122] Suspended particles carried away from the bottle by the gas-liquid flow must be effectively collected to prevent secondary sedimentation. The system deploys an electrostatic adsorption device (EAD) at the top of the workstation. This device consists of two sets of parallel electrode plates: the upper electrode is a tungsten wire mesh, subjected to a +15 kV high-voltage direct current; the lower electrode is a stainless steel dust collection plate, grounded and circulated with cooling water. When the gas-liquid flow containing debris passes through the electrode gap, the particles ionize and become charged (mainly negatively charged) in the high-voltage electric field, and are driven directionally towards the dust collection plate by Coulomb force. The electric field strength is designed to be 5 kV / cm, ensuring a capture efficiency of >99% for particles larger than 1 micrometer. The surface of the dust collection plate is coated with Teflon, and accumulated dust is automatically removed by periodic mechanical scraping (60-second cycle), with the dust falling into a sealed waste bin.

[0123] After the entire cleaning process is completed, the system initiates a final evaluation: 1) Surface roughness is inspected using a laser confocal microscope (Ra value must be <0.05 micrometers); 2) The water contact angle is measured using a contact angle meter (<5 degrees to prove no oil contamination); 3) A white glove wiping test shows no visible residue. Once the standards are met, the robotic arm transfers the glass bottle to the next process; at this point, the workpiece is called a semi-finished glass bottle blank (SFGB). This semi-finished blank has no macroscopic scratches, no attached debris, and no organic contamination on its surface. Its microscopic roughness is reduced by 90% compared to the rough blank, creating an ideal substrate for subsequent nanoscale polishing. All process parameters (such as UV dose, electrostatic voltage, and cleaning time) are recorded in the MES (Manufacturing Execution System) to achieve full-process quality traceability.

[0124] S204 involves placing the finely polished glass bottle preform into a sealed cavity, spraying it with a suspension of nano-sized silica flexible abrasive particles, and then performing final polishing with pulsed ultrasonic vibration. Through the synergistic effect of particle Brownian motion and ultrasonic cavitation, residual micron-sized contaminants are removed, resulting in a finished cosmetic glass bottle with a clean surface and a preset gloss level.

[0125] Specifically, silica particles of a specific size can be dispersed in deionized water, and an appropriate amount of polyvinylpyrrolidone can be added as a dispersant. A stable suspension can be prepared by ultrasonic homogenization and used as the grinding medium for final polishing.

[0126] The core of this step is the preparation of a highly stable polishing medium with a precise nanoscale particle size distribution. First, silica particles of a specific size are rigorously screened, with an average particle size controlled within the range of 50 to 100 nanometers (e.g., 80 nanometers), and a particle size distribution variation coefficient of less than 15%. This particle size selection is based on two considerations: first, it must be smaller than the visible light wavelength (approximately 400-700 nanometers) to avoid optical scattering during polishing that could affect monitoring; second, it must possess sufficient Brownian motion activity to penetrate microscopic depressions. The selected silica particles undergo hydroxylation surface treatment to enrich their surface with silanol groups, enhancing hydrophilicity. The resistivity of deionized water must reach above 18 megohms·cm to ensure dispersion stability without electrolyte interference. Polyvinylpyrrolidone (PVP) is used as a dispersant, with a molecular weight range of 40,000 to 60,000 Daltons (e.g., 50,000 Daltons), and its concentration is precisely controlled at 0.5% to 0.8% by mass (e.g., 0.65%). PVP's pyrrolidone rings are adsorbed onto the silica surface by van der Waals forces, and its long molecular chains extend in water to form a steric hindrance layer, effectively resisting particle aggregation.

[0127] The ultrasonic homogenization process employs a staged energy input strategy. The first stage involves premixing silica particles, PVP, and deionized water, then placing the mixture in a titanium alloy amplitude transformer ultrasonic processor for five minutes of intense dispersion at a frequency of 20 kHz and a power of 1500 W. The high-intensity cavitation effect in this stage breaks down primary aggregates, while the acoustic flow effect promotes macroscopic mixing. The second stage involves switching to a frequency of 40 kHz and a power of 800 W for thirty minutes of gentle homogenization. In this stage, the cavitation intensity is controlled by adjusting the ultrasonic duty cycle to 50% (0.5 seconds on / 0.5 seconds off), focusing on optimizing the adsorption layer structure on the particle surface and eliminating microbubbles. The entire process is conducted in a constant-temperature water bath (25°C ± 1°C) to prevent localized overheating that could lead to PVP molecular chain degradation. The final stable suspension must meet three criteria: an absolute Zeta potential greater than 35 mV (e.g., -42 mV), characterizing electrostatic stability; a polydispersity index of less than 0.15 as measured by dynamic light scattering, characterizing homogeneity; and a sedimentation volume ratio of less than 5% after standing for 24 hours, characterizing anti-settling properties. This suspension is stored in a nitrogen-protected, sealed container and has a shelf life of 72 hours.

[0128] To ensure the traceability and batch consistency of the grinding media, an online quality monitoring system was established. An ultraviolet-visible spectrophotometer (detecting absorbance at 600 nm wavelength and correlated with particle concentration) and a conductivity meter (monitoring ionic contamination) were integrated into the suspension delivery pipeline. When absorbance fluctuations exceed 3% of the baseline value or conductivity exceeds 5 μSiemens / cm, the system automatically triggers an alarm and suspends feeding. Each batch of suspension requires random sampling using a scanning electron microscope to verify particle morphology (sphericity greater than 0.9) and the absence of hard agglomeration. This strictly controlled nano-grinding media possesses three functions: flexible particles achieve micro-cutting through rolling friction; Brownian motion drives particles into submicron-level defects; and surface active sites adsorb residual contaminants.

[0129] A pressure-type atomizing nozzle is used to spray nano-grinding media onto the surface of a glass bottle in a fan-shaped spray pattern. The spray particle size distribution is monitored in real time by a laser particle size analyzer, and the spray pressure is dynamically adjusted to form a uniform liquid film.

[0130] The spray system employs a multi-stage precision control architecture. The pressure-type atomizing nozzles feature a wide-angle fan-shaped design, with the spray angle precisely set to 80 degrees ± 2 degrees to ensure coverage of the curved surface of the glass bottle. The nozzles integrate a vortex chamber and a precision slit (200 micrometers wide), controlling the liquid film breakup mode by adjusting the inlet pressure. The grinding suspension is delivered by a high-pressure plunger pump, with an operating pressure range set from 0.5 MPa to 3 MPa (initial value 1.2 MPa). A key innovation lies in dual-fluid assisted atomization: the main channel delivers the grinding fluid, while a ring-shaped secondary channel introduces dry nitrogen (99.999% purity), at 80% of the main pressure (e.g., 0.96 MPa). The laminar flow of nitrogen surrounds the atomizing cone, suppressing spray drift and preventing water vapor condensation from affecting the polishing environment. The nozzle array is equidistantly distributed according to the three-dimensional contour of the glass bottle, with the spacing dynamically adjusted according to the bottle diameter (e.g., 25 mm spacing for a 60 mm diameter bottle).

[0131] The Laser Diffraction Analyzer (LDA) real-time monitoring system is the core guarantee for the uniformity of the liquid film. This instrument uses a helium-neon laser light source (wavelength 632.8 nm), and its detector array covers a measurement range from 0.1 μm to 1000 μm. A detection window is set at 150 mm from the nozzle outlet, acquiring spray data ten times per second. The analysis software calculates key parameters in real time: volume average particle size (D50), span value ((D90-D10) / D50), and the proportion of coarse droplets larger than 5 μm. When D50 deviates from the target value (e.g., target 75 μm) by ±5% or the proportion of coarse droplets exceeds 0.1%, a dynamic adjustment mechanism is triggered. The adjustment logic is based on fuzzy proportional-integral control: if D50 is too large, indicating insufficient atomization, the system increases the plunger pump pressure in steps at a rate of 0.05 MPa per second; if the span value increases, indicating a widening of the particle size distribution, the nitrogen-assisted pressure is increased simultaneously by 3% to 5%.

[0132] To form a uniform liquid film, three core indicators must be met: film thickness uniformity error less than ±10% (target thickness 15 micrometers), coverage greater than 99.8%, and no dry spots or liquid accumulation areas. The implementation strategy includes: First, based on the coordinated control of the glass bottle rotation speed (e.g., 10 rpm) and nozzle movement speed (e.g., 20 mm / s), a motion controller calculates the spatial trajectory to ensure the overlap spraying rate remains at 30%. Second, hydrophobic micro-area positioning marks are pre-placed on the bottle surface, and the liquid film spreading process is captured by a high-speed camera (1000 frames / second). If an area with a contact angle exceeding 90 degrees is detected, the nozzle pressure in that area is increased by 15%. Third, an infrared thermal imager monitors the liquid film evaporation state (temperature resolution 0.1 degrees Celsius), and when the local temperature difference exceeds 0.5 degrees Celsius, supplementary spraying is triggered. This dynamic closed-loop system enables nanoparticles to form a molecular-level lubricating film with self-healing properties on the glass surface.

[0133] Pulsed ultrasonic vibration with a specific pulse vibration frequency is applied to a glass bottle covered with a uniform liquid film. The generation and collapse cycle of cavitation bubbles are controlled by duty cycle modulation. Combined with fluid dynamics simulation, the transducer position is optimized to enhance the selective polishing effect.

[0134] The pulsed ultrasonic vibration system consists of three main modules: a piezoelectric ceramic transducer array (PZT-8 material), a pulse power amplifier, and a digital signal generator. The specific pulse vibration frequency is selected as 28 kHz ± 500 Hz. This frequency, after modal analysis, avoids the natural resonant frequency of the glass bottle (typically 18-22 kHz) to prevent structural damage. The pulse waveform uses rectangular modulation, with key parameters including: pulse width (Ton) ranging from 50 to 500 microseconds, pulse interval (Toff) ranging from 100 to 1000 microseconds, and peak acoustic power density of 0.5 to 3 W / cm² (initial value 1.8 W / cm²). The transducer transmits the vibration to the bottom plate of the polishing tank via a hydraulic coupling rod, generating longitudinal waves in the liquid film. Frequency stability is controlled by a phase-locked loop circuit, with phase jitter less than 0.1 degrees.

[0135] Duty cycle modulation is the core technology for controlling cavitation behavior. The duty cycle is defined as Ton / (Ton+Toff), with an initial value of 40%. This is achieved by adjusting the ratio of Ton to Toff: when Ton is extended to 300 microseconds, it promotes stable growth of cavitation bubbles, enhances the impact force of the microjets, and is suitable for removing deeply embedded contaminants; when Toff is extended to 800 microseconds, it prolongs the recovery period after bubble collapse, facilitating the redistribution of abrasive particles. More advanced modulation employs a chaotic sequence algorithm: using a Lorentz attractor model to generate a non-periodic duty cycle variation sequence, causing cavitation bubbles to be randomly distributed in the spatiotemporal domain, eliminating polishing unevenness caused by standing waves. The cavitation intensity is monitored in real time by a hydrophone array (center frequency 28 kHz, sensitivity -180 dB). When the sound pressure level exceeds 0.15 MPa, the duty cycle is automatically reduced by 5%, preventing the formation of micro-pits on the glass surface.

[0136] Fluid dynamics simulation optimization was performed using finite element analysis software to conduct three-dimensional transient simulations. The model included the glass bottle geometry, liquid film boundary layer, nanoparticle two-phase flow, and ultrasonic pressure field. Key optimization objectives were: to achieve sound pressure focusing (enhancing by more than 2.5 times) in the surface protrusion area and to maintain laminar flow protection in the depression area. Simulation results revealed that the transducers should be arranged in a helical array (e.g., six groups spaced at 60 degrees), with an optimal distance from the bottle wall of one-quarter wavelength (approximately 13 mm in water); setting the tilt angle to 15 degrees can suppress surface wave interference. Experimental verification employed particle image velocimetry: fluorescent tracer particles (5 μm) were added to the liquid film, illuminated by a dual-pulse laser sheet, and the flow field was captured by a high-speed camera. Optimized data showed that the vorticity in the micro-protrusion area increased threefold, the shear stress reached 12 Pascals, while the flow velocity in the depression area was controlled below 0.2 m / s. This selective polishing effect reduced the arithmetic mean deviation (Ra) of the surface profile from 0.3 μm to 0.02 μm.

[0137] The polishing effect is monitored in real time by an online gloss meter. The ultrasonic power and nanoparticle concentration are dynamically adjusted based on the monitoring data. When the gloss reaches the preset standard, the cleaning program is triggered. The polishing media is recycled through a circulating filtration system to obtain a finished cosmetic glass bottle with a clean surface and the preset gloss.

[0138] The online gloss meter employs a 60-degree geometric measurement angle (international standard ISO 2813) and scans twenty points per second. The optical system includes a halogen light source (color temperature 2856K), an interference filter (peak wavelength 560 nm), and a silicon photodetector. The instrument contacts the bottle surface during the polishing gap (Toff period) via a pneumatic lifting mechanism, and the measuring head applies a constant contact force of 0.5 Newtons to ensure repeatability. After noise reduction via Kalman filtering, the monitoring data is used to calculate two key values: specular reflectance (Gs, unit GU) and haze (characterizing the proportion of scattered light). Preset gloss standards are set according to customer requirements (e.g., high-end perfume bottles require Gs ≥ 95 GU and Haze ≤ 2%). The system establishes a gloss-roughness mapping database: when Ra = 0.02 micrometers, Gs = 97 GU; for every 0.01 micrometer increase in Ra, Gs decreases by 6 GU.

[0139] The dynamic adjustment strategy employs multivariate decoupled control: First, when the real-time Gs value is lower than the target trajectory (e.g., it should reach 80 GU in the 3rd minute but is actually measured at 75 GU), the ultrasonic power is increased in steps of 0.2 W / cm², while simultaneously detecting the cavitation noise spectrum. If a sudden increase in energy occurs in the high-frequency band (>100 kHz) of 1 / 3 octave band, the pressurization is stopped. Second, when the haze value exceeds the standard, it indicates the presence of micro-scratches or particle residues on the surface. The system increases the nanoparticle concentration from the initial 5% mass fraction to 7%, and simultaneously increases the PVP dispersant ratio to 0.8%. Third, a polishing efficiency prediction model is established: if the Gs increase rate per unit time is less than 0.3 GU per second, the ultrasonic duty cycle is increased by 10% and the spray flow rate by 15%. All adjustment commands are transmitted to the actuator via the OPC protocol, with a response delay of less than 100 milliseconds.

[0140] The circulating filtration and finished product output process involves four precise steps: First, when three consecutive measurements by the online gloss meter fall within the acceptable range (e.g., Gs=96GU±0.5GU), an audible and visual alarm is triggered, and the cleaning program is initiated. Second, the circulating filtration system immediately switches valves: the polishing liquid flows into a multi-stage series filter—the first stage uses a stainless steel sintered filter element (5-micron pore size) to intercept large particles; the second stage uses a hollow fiber ultrafiltration membrane (100,000 Dalton molecular weight cutoff) to remove aggregates; and the final stage uses an electrodialysis desalination module to maintain water quality. The recovery rate is greater than 95%, and the filtrate is reused after replenishing with fresh polishing liquid. Third, the glass bottles undergo three stages of cleaning: ultrasonic rinsing with 80°C deionized water (40 kHz, 3 minutes); anhydrous ethanol vapor degreasing (50°C, 2 minutes); and negative ion air knife drying (30 m / s). Fourth, the finished products pass through an automated sorting line: a machine vision system re-inspects surface defects (2-micron resolution), a laser interferometer checks surface shape accuracy (λ / 10), and qualified products are coded and packaged. The final cosmetic glass bottle achieves a surface cleanliness level of ISO 14644-1 Class 5, with a gloss fluctuation range of ±1GU, realizing an optical surface with nanometer-level precision.

[0141] As can be seen, by modifying the base material of the abrasive with gradient mesh size, and then using laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface, a composite abrasive polishing part is formed. The composite abrasive polishing part is assembled on the execution end of a six-axis linkage robotic arm, and differentiated polishing is performed in different areas based on the image recognition results of the surface contaminant distribution density, resulting in a rough glass bottle blank after preliminary polishing. On the surface of the rough glass bottle blank, an inert gas flow with a microbubble generator is used for synchronous blowing, while the composite abrasive polishing part is controlled to vibrate slightly, resulting in a semi-finished glass bottle blank. The semi-finished glass bottle blank is placed in a sealed cavity for spraying, and then polished with pulsed ultrasonic vibration to obtain the finished cosmetic glass bottle. This enables high-precision adaptive polishing of the curved surface of the glass bottle, improving surface cleanliness and gloss uniformity.

[0142] Another embodiment of the present invention provides a surface polishing system for cosmetic glass bottles, see [link to relevant documentation]. Figure 3 The system may include:

[0143] The cutting module 301 is used to perform gradient mesh abrasive composite modification on the sandpaper base material. It uses laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface, forming a composite sandpaper polishing part that precisely fits the curved surface of the bottle.

[0144] The polishing module 302 is used to assemble the composite sandpaper polishing parts onto the execution end of the six-axis linkage robotic arm. It acquires three-dimensional morphological data of the glass bottle surface through infrared scanning, dynamically adjusts the rotation speed of the servo motor and the contact pressure of the polishing parts based on the morphological data, and performs zoned differentiated polishing in combination with the image recognition results of the surface contaminant distribution density to obtain the glass bottle blank after preliminary polishing.

[0145] The purging module 303 is used to simultaneously purge the surface of the glass bottle blank after preliminary polishing with an inert gas flow equipped with a microbubble generator, while controlling the composite sandpaper polishing parts to vibrate slightly. Through gas-solid coupling, residual polishing debris is removed to obtain a finely polished glass bottle semi-finished blank.

[0146] The spray module 304 is used to place the finely polished glass bottle semi-finished blank into a sealed cavity, and spray it with a suspension of nano-sized silica flexible abrasive particles. Combined with pulsed ultrasonic vibration, it performs final polishing. Through the synergistic effect of particle Brownian motion and ultrasonic cavitation, residual micron-sized contaminants are removed, resulting in a cosmetic glass bottle with a clean surface and a preset gloss level.

[0147] This invention also provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0148] Specifically, in this embodiment, the storage medium can be configured to store a computer program for performing the following steps:

[0149] S201 involves modifying the base material of sandpaper with gradient abrasive composite, and using laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface to form a composite sandpaper polishing part that precisely fits the curved surface of the bottle.

[0150] S202, the composite sandpaper grinding part is assembled on the execution end of the six-axis linkage robotic arm. The three-dimensional morphological data of the glass bottle surface is obtained by infrared scanning. Based on the morphological data, the rotation speed of the servo motor and the contact pressure of the grinding part are dynamically adjusted. Combined with the image recognition results of the surface contaminant distribution density, the partitioned differential grinding is carried out to obtain the glass bottle blank after preliminary grinding.

[0151] S203, on the surface of the glass bottle blank after preliminary polishing, an inert gas flow with a microbubble generator is used for synchronous purging, while the composite sandpaper polishing parts are controlled to vibrate slightly. The residual polishing debris is removed through gas-solid coupling, and a semi-finished glass bottle blank after fine polishing is obtained.

[0152] S204 involves placing the finely polished glass bottle preform into a sealed cavity, spraying it with a suspension of nano-sized silica flexible abrasive particles, and then performing final polishing with pulsed ultrasonic vibration. Through the synergistic effect of particle Brownian motion and ultrasonic cavitation, residual micron-sized contaminants are removed, resulting in a finished cosmetic glass bottle with a clean surface and a preset gloss level.

[0153] This invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0154] Specifically, the aforementioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the aforementioned processor, and the input / output device is connected to the aforementioned processor.

[0155] Specifically, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0156] S201 involves modifying the base material of sandpaper with gradient abrasive composite, and using laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface to form a composite sandpaper polishing part that precisely fits the curved surface of the bottle.

[0157] S202, the composite sandpaper grinding part is assembled on the execution end of the six-axis linkage robotic arm. The three-dimensional morphological data of the glass bottle surface is obtained by infrared scanning. Based on the morphological data, the rotation speed of the servo motor and the contact pressure of the grinding part are dynamically adjusted. Combined with the image recognition results of the surface contaminant distribution density, the partitioned differential grinding is carried out to obtain the glass bottle blank after preliminary grinding.

[0158] S203, on the surface of the glass bottle blank after preliminary polishing, an inert gas flow with a microbubble generator is used for synchronous purging, while the composite sandpaper polishing parts are controlled to vibrate slightly. The residual polishing debris is removed through gas-solid coupling, and a semi-finished glass bottle blank after fine polishing is obtained.

[0159] S204 involves placing the finely polished glass bottle preform into a sealed cavity, spraying it with a suspension of nano-sized silica flexible abrasive particles, and then performing final polishing with pulsed ultrasonic vibration. Through the synergistic effect of particle Brownian motion and ultrasonic cavitation, residual micron-sized contaminants are removed, resulting in a finished cosmetic glass bottle with a clean surface and a preset gloss level.

[0160] The above description, based on the embodiments shown in the figures, details the structure, features, and effects of the present invention. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A method for polishing the surface of a cosmetic glass bottle, characterized in that, The method includes: The base material of sandpaper is modified with gradient abrasive composite, and three-dimensional cutting is performed according to the curved contour of the glass bottle surface using laser cutting technology to form a composite sandpaper polishing part that precisely fits the curved surface of the bottle. The composite sandpaper polishing part is assembled on the execution end of a six-axis linkage robotic arm. The three-dimensional morphological data of the glass bottle surface is obtained by infrared scanning. Based on the morphological data, the rotation speed of the servo motor and the contact pressure of the polishing part are dynamically adjusted. Combined with the image recognition results of the surface contaminant distribution density, zoned differential polishing is carried out to obtain the glass bottle blank after preliminary polishing. On the surface of the glass bottle blank after preliminary polishing, an inert gas flow with a microbubble generator is used for synchronous purging, while the composite sandpaper polishing parts are controlled to vibrate slightly. The residual polishing debris is removed through gas-solid coupling, resulting in a finely polished glass bottle semi-finished blank. The finely polished glass bottle preform is placed in a sealed cavity and sprayed with a suspension of nano-sized silica flexible abrasive particles. Finally, it is polished by pulsed ultrasonic vibration. The residual micron-sized contaminants are removed through the synergistic effect of particle Brownian motion and ultrasonic cavitation, resulting in a cosmetic glass bottle with a clean surface and a preset gloss level.

2. The method according to claim 1, characterized in that, The process involves modifying the base material of the abrasive with gradient mesh size, and then using laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface, forming a composite abrasive polishing part that precisely fits the curved surface of the bottle. This includes: Natural sand-like material is placed in supercritical carbon dioxide fluid, and the fluid density is controlled by adjusting the pressure and temperature to achieve the regulation of the nanoscale pore structure of the sand-like material fibers, forming a porous carrier structure, which provides anchoring points for subsequent abrasive particle adhesion. Electrophoretic deposition technology was used to sequentially deposit 800-mesh, 1500-mesh, and 3000-mesh zirconia abrasive grains on the surface of a porous carrier structure. The abrasive grain deposition rate was controlled by pulse current density to form a gradient abrasive grain layer, which constitutes a gradient abrasive composite substrate. Using a femtosecond laser micro-nano processing system, contour tracking and cutting of a gradient abrasive composite substrate is performed based on the three-dimensional model data of a cosmetic glass bottle. At the same time, micro-serrated joints are generated at the cutting edge to complete the preparation of a three-dimensional pre-cut and polished part. The three-dimensional pre-cut polishing part is immersed in a silane coupling agent solution for surface activation, and then hot-pressed with a shape memory polymer substrate. The shape memory effect of the substrate gives the polishing part an adaptive bonding ability, forming a composite sandpaper polishing part.

3. The method according to claim 2, characterized in that, The composite sandpaper polishing component is assembled onto the execution end of a six-axis linkage robotic arm. Three-dimensional morphological data of the glass bottle surface is acquired through infrared scanning. Based on this morphological data, the rotational speed of the servo motor and the contact pressure of the polishing component are dynamically adjusted. Combined with the image recognition results of surface contaminant distribution density, differentiated polishing is performed in different zones to obtain a preliminary polished glass bottle blank, including: An infrared thermal imager and a structured light scanner are integrated into the execution end of the robotic arm to perform multimodal scanning on the surface of the glass bottle. Subsurface defects are identified through thermal wave imaging technology, and the scan data is fused to generate a three-dimensional point cloud dataset containing morphology and defect information. Based on a 3D point cloud dataset, a standard bottle model is matched using a dynamic time warping algorithm to calculate a local curvature deviation map. Combined with a deep learning semantic segmentation model, the types and distribution densities of surface contaminants are identified, and a differentiated polishing strategy map is constructed. Using curvature deviation map and pollutant distribution density as input, the servo motor speed and contact pressure parameters are extrapolated in real time through fuzzy control rule base to form a dynamic control parameter set; Based on the dynamic control parameter set, the six-axis robotic arm is driven to move along a preset trajectory and simultaneously perform zoned grinding operations. The dynamic error of the robotic arm is corrected through iterative learning control algorithms to complete the initial grinding and obtain the glass bottle blank.

4. The method according to claim 3, characterized in that, The process involves simultaneously purging the surface of the pre-polished glass bottle blank with an inert gas stream equipped with a microbubble generator, while simultaneously controlling the composite abrasive grinding component to vibrate micro-vibrate. This gas-solid coupling effect removes residual grinding debris, resulting in a finely polished semi-finished glass bottle blank. The process includes: Argon gas is introduced into a venturi tube and mixed with deionized water. Microbubbles of a specific diameter are generated at the gas-liquid interface through ultrasonic cavitation effect, forming a gas-liquid two-phase flow, which serves as a medium for debris removal. The distribution of debris on the surface of the glass bottle is monitored in real time by a high-speed camera. Particle image velocimetry technology is used to calculate the trajectory of the debris. Based on the trajectory data, the injection angle and velocity of the gas-liquid two-phase flow are dynamically adjusted to form precise purging parameters. Based on the purging parameters, the composite sandpaper grinding part is controlled to perform elliptical trajectory micro-vibration at a specific micro-vibration frequency. The amplitude and phase difference are optimized by vibration mode analysis to enhance the chip resonance detachment effect. Ultraviolet radiation is introduced during the purging process to decompose organic residues through photocatalytic oxidation. Suspended particles are collected by an electrostatic adsorption device to complete the removal of residual debris and obtain a semi-finished glass bottle.

5. The method according to claim 4, characterized in that, The process involves placing a finely polished semi-finished glass bottle into a sealed cavity, spraying it with a suspension of nano-sized silica flexible abrasive particles, and then performing final polishing using pulsed ultrasonic vibration. Through the synergistic effect of particle Brownian motion and ultrasonic cavitation, residual micron-sized contaminants are removed, resulting in a finished cosmetic glass bottle with a clean surface and a preset gloss level. Silica particles of a specific size are dispersed in deionized water, and an appropriate amount of polyvinylpyrrolidone is added as a dispersant. A stable suspension is prepared by ultrasonic homogenization and used as the grinding medium for final polishing. A pressure-type atomizing nozzle is used to spray nano-grinding media onto the surface of a glass bottle in a fan-shaped spray pattern. The spray particle size distribution is monitored in real time by a laser particle size analyzer, and the spray pressure is dynamically adjusted to form a uniform liquid film. Pulsed ultrasonic vibration with a specific pulse vibration frequency is applied to a glass bottle covered with a uniform liquid film. The generation and collapse cycle of cavitation bubbles are controlled by duty cycle modulation. Combined with fluid dynamics simulation, the transducer position is optimized to enhance the selective polishing effect. The polishing effect is monitored in real time by an online gloss meter. The ultrasonic power and nanoparticle concentration are dynamically adjusted based on the monitoring data. When the gloss reaches the preset standard, the cleaning program is triggered. The polishing media is recycled through a circulating filtration system to obtain a finished cosmetic glass bottle with a clean surface and the preset gloss.

6. A surface polishing system for cosmetic glass bottles, characterized in that, The system includes: The cutting module is used to modify the sandpaper base material with gradient abrasive composite. It uses laser cutting technology to perform three-dimensional cutting according to the curved contour of the glass bottle surface, forming a composite sandpaper polishing part that precisely fits the curved surface of the bottle. The polishing module is used to assemble the composite sandpaper polishing parts onto the execution end of the six-axis linkage robotic arm. It acquires three-dimensional morphological data of the glass bottle surface through infrared scanning, dynamically adjusts the rotation speed of the servo motor and the contact pressure of the polishing parts based on the morphological data, and performs zoned differentiated polishing in combination with the image recognition results of the surface contaminant distribution density to obtain the glass bottle blank after preliminary polishing. The purging module is used to simultaneously purge the surface of the glass bottle blank after preliminary polishing with an inert gas flow equipped with a microbubble generator, while controlling the composite sandpaper polishing parts to vibrate slightly. Through gas-solid coupling, residual polishing debris is removed to obtain a finely polished glass bottle semi-finished blank. The spray module is used to place the finely polished glass bottle semi-finished blank into a sealed cavity, and spray it with a suspension of nano-sized silica flexible abrasive particles. Combined with pulsed ultrasonic vibration, the final polishing is carried out. The residual micron-sized contaminants are removed through the synergistic effect of particle Brownian motion and ultrasonic cavitation, resulting in a cosmetic glass bottle with a clean surface and a preset gloss level.

7. The system according to claim 6, characterized in that, The cropping module is specifically used for: Natural sand-like material is placed in supercritical carbon dioxide fluid, and the fluid density is controlled by adjusting the pressure and temperature to achieve the regulation of the nanoscale pore structure of the sand-like material fibers, forming a porous carrier structure, which provides anchoring points for subsequent abrasive particle adhesion. Electrophoretic deposition technology was used to sequentially deposit 800-mesh, 1500-mesh, and 3000-mesh zirconia abrasive grains on the surface of a porous carrier structure. The abrasive grain deposition rate was controlled by pulse current density to form a gradient abrasive grain layer, which constitutes a gradient abrasive composite substrate. Using a femtosecond laser micro-nano processing system, contour tracking and cutting of a gradient abrasive composite substrate is performed based on the three-dimensional model data of a cosmetic glass bottle. At the same time, micro-serrated joints are generated at the cutting edge to complete the preparation of a three-dimensional pre-cut and polished part. The three-dimensional pre-cut polishing part is immersed in a silane coupling agent solution for surface activation, and then hot-pressed with a shape memory polymer substrate. The shape memory effect of the substrate gives the polishing part an adaptive bonding ability, forming a composite sandpaper polishing part.

8. The system according to claim 7, characterized in that, The polishing module is specifically used for: An infrared thermal imager and a structured light scanner are integrated into the execution end of the robotic arm to perform multimodal scanning on the surface of the glass bottle. Subsurface defects are identified through thermal wave imaging technology, and the scan data is fused to generate a three-dimensional point cloud dataset containing morphology and defect information. Based on a 3D point cloud dataset, a standard bottle model is matched using a dynamic time warping algorithm to calculate a local curvature deviation map. Combined with a deep learning semantic segmentation model, the types and distribution densities of surface contaminants are identified, and a differentiated polishing strategy map is constructed. Using curvature deviation map and pollutant distribution density as input, the servo motor speed and contact pressure parameters are extrapolated in real time through fuzzy control rule base to form a dynamic control parameter set; Based on the dynamic control parameter set, the six-axis robotic arm is driven to move along a preset trajectory and simultaneously perform zoned grinding operations. The dynamic error of the robotic arm is corrected through iterative learning control algorithms to complete the initial grinding and obtain the glass bottle blank.

9. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the method of any one of claims 1-5 when it is run.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method of any one of claims 1-5.

Images