Method and apparatus for applying a surface fungicide to a glass article

Abstract

The application provides a glass product surface mildew-proof powder spreading method and equipment to realize uniform spreading of mildew-proof powder on the surface of the glass product. The method comprises the following steps: obtaining depth data and charge distribution data of the surface of the glass product; calculating the curvature data of the surface of the glass product according to the depth data; performing finite element grid division on the surface of the glass product according to the curvature data to obtain a plurality of grids, and calculating the target electric field intensity of each grid; determining the first voltage of each grid according to the target electric field intensity of each grid; determining the second voltage of each grid according to the charge distribution data; calculating the difference between the first voltage and the second voltage of each grid to obtain the third voltage; and spreading the mildew-proof powder on each region of the surface of the glass product in a first spreading mode, wherein when the mildew-proof powder spreader spreads the mildew-proof powder on the target region of the surface of the glass product, the voltage of each electrode of the electrostatic generator is controlled to be the third voltage of the target grid of the electrode at present.

Description

Technical Field

[0001] This invention relates to the field of anti-mildew treatment technology for glass product surfaces, and in particular to a method and equipment for applying anti-mildew powder to the surface of glass products. Background Technology

[0002] When glass products come into contact with moisture, dust, and other substances in the air, the dust adsorbed on the surface can easily lead to mold growth. Therefore, certain measures are usually taken to protect the surface of glass products during packaging, transportation, and storage.

[0003] Currently, glass products can be protected by spraying anti-mold powder onto their surfaces to prevent mold growth caused by dust adhering to the surface. This is primarily done manually, or via a powder spraying machine that automatically applies the powder without any special spraying control.

[0004] However, due to the unevenness of the glass surface, the application of anti-mold powder without special control can lead to uneven application. This results in some areas of the glass surface with too little anti-mold powder being poorly protected against mold growth, while some areas with too much anti-mold powder are wasting the powder.

[0005] In summary, existing technologies suffer from uneven application of anti-mildew powder to the surface of glass products. Summary of the Invention

[0006] This invention provides a method and equipment for spreading anti-mildew powder on the surface of glass products, so as to achieve uniform spreading of anti-mildew powder on the surface of glass products.

[0007] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for applying anti-mildew powder to the surface of glass products, comprising:

[0008] Acquire depth and charge distribution data on the surface of glass products;

[0009] The curvature data of the glass product surface is calculated based on the depth data;

[0010] The surface of the glass product is divided into multiple meshes using finite element methods based on the curvature data, and the target electric field intensity of each mesh is calculated.

[0011] The first voltage corresponding to each grid is determined based on the target electric field intensity corresponding to each grid.

[0012] The second voltage corresponding to each grid is determined based on the charge distribution data;

[0013] For each grid cell, the difference between the first voltage and the second voltage is calculated to obtain the third voltage.

[0014] Anti-mildew powder is spread onto various areas of the surface of the glass product using a first spreading method. The first spreading method is as follows: when the anti-mildew powder spreader spreads anti-mildew powder onto the target area of ​​the glass product surface, the voltage of each electrode of the electrostatic generator is controlled to be the third voltage of the target grid corresponding to the electrode, so that when the anti-mildew powder is spread to the corresponding target grid through the corresponding electrode, it carries electrostatic charge and is adsorbed on the target grid. The target grid is the grid contained in the target area.

[0015] The electrostatic generator is located at the spraying port of the anti-mildew powder applicator, and the electrodes of the electrostatic generator are mutually insulated.

[0016] Secondly, the present invention also provides a controller for applying anti-mildew powder to the surface of glass products, comprising:

[0017] The measurement module is used to acquire depth data and charge distribution data on the surface of glass products;

[0018] The curvature calculation module is used to calculate the curvature data of the glass product surface based on the depth data.

[0019] The finite element calculation module is used to divide the surface of the glass product into multiple meshes based on the curvature data, and to calculate the target electric field intensity of each mesh.

[0020] The first voltage calculation module is used to determine the first voltage corresponding to each grid according to the target electric field intensity corresponding to each grid.

[0021] The second voltage calculation module is used to determine the second voltage corresponding to each grid based on the charge distribution data;

[0022] The third voltage calculation module is used to calculate the difference between the corresponding first voltage and the corresponding second voltage for each grid, so as to obtain the third voltage;

[0023] The first powder-spreading module is used to spread anti-mildew powder to various areas of the surface of the glass product in a first spreading method. The first spreading method is as follows: when the anti-mildew powder spreader spreads anti-mildew powder to the target area on the surface of the glass product, it controls the voltage of each electrode of the electrostatic generator to be the third voltage of the target grid corresponding to the electrode, so that when the anti-mildew powder is spread to the corresponding target grid through the corresponding electrode, it carries electrostatic charge and is adsorbed on the target grid. The target grid is the grid contained in the target area.

[0024] The electrostatic generator is located at the spraying port of the anti-mildew powder applicator, and the electrodes of the electrostatic generator are mutually insulated.

[0025] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the method for applying anti-mildew powder to the surface of glass products as described in the first aspect.

[0026] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the method for applying anti-mildew powder to the surface of glass products as described in the first aspect.

[0027] Fifthly, the present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being used to implement the method of applying anti-mildew powder to the surface of glass products as described in the first aspect when the computer is run.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] When applying anti-mildew powder to the surface of glass products, this invention adaptively adjusts the voltage of the electrodes of the electrostatic generator that applies electrostatic charge to the anti-mildew powder according to the shape of the glass product surface and the electrostatic charge carried by the glass product. This allows the anti-mildew powder applied at different locations on the glass product surface to be adsorbed based on the corresponding electrostatic charge, ultimately achieving uniform application of the anti-mildew powder to the glass product surface. This avoids both insufficient application of anti-mildew powder in some areas, which would result in ineffective anti-mildew treatment, and excessive application of anti-mildew powder in some areas, which would lead to waste. Attached Figure Description

[0030] Figure 1 This is a cross-sectional side view of the anti-mildew powder spreading machine for glass products provided in an embodiment of the present invention.

[0031] Figure 2 This is a front view of the electrostatic generator of the anti-mildew powder spreading machine for glass products provided in an embodiment of the present invention.

[0032] Figure 3 This is a schematic flowchart of the method for applying anti-mildew powder to the surface of glass products according to an embodiment of the present invention.

[0033] Figure 4 This is a schematic diagram of the training steps of a fully connected neural network provided in an embodiment of the present invention.

[0034] Figure 5This is a schematic diagram of the structure of the anti-mildew powder application controller for glass products provided in an embodiment of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] This invention provides a method for applying anti-mildew powder to the surface of glass products. The method can be applied to an anti-mildew powder application machine for glass products, where an electrostatic generator is positioned at the application port of the anti-mildew powder applicator, and the electrodes of the electrostatic generator are mutually insulated.

[0037] In one implementation, refer to Figure 1 The glass product surface anti-mildew powder spreading machine (100) includes a controller (not shown in the figure), a frame (101), a glass product conveying mechanism (102), an anti-mildew powder storage device (103), an anti-mildew powder spreader (104), an electrostatic generator (105), a charge induction array sensor (not shown in the figure), and a positioning sensor (not shown in the figure). Wherein:

[0038] A glass product conveying mechanism (102) is mounted on a frame (101) and is used to move the glass products (200) under the control of a controller, so that each position of the glass product passes through the spraying port of the anti-mildew powder spreader (104) in sequence. Specifically, the glass product conveying mechanism (102) can be implemented using a conveyor belt, rollers, or other structures.

[0039] An anti-mold powder storage device (103) is installed above the glass product conveying mechanism (102) for storing anti-mold powder, wherein the outlet of the anti-mold powder storage device (103) is connected to the anti-mold powder sprinkler (104).

[0040] An anti-mold powder applicator (104) is used to apply anti-mold powder to the surface of glass products (200). In one embodiment, reference... Figure 1 The anti-mold powder applicator (104) is a centrifugal fan-blade applicator. The anti-mold powder applicator (104) can rotate through the rotating fan blade structure, and the anti-mold powder in the applicator (104) is sprayed onto the surface of the glass product (200) by centrifugal force through the spraying port of the anti-mold powder applicator (104). In one embodiment, the anti-mold powder applicator (104) is a jet-type applicator. The anti-mold powder applicator (104) can spray the anti-mold powder in the applicator (104) from the spraying port onto the surface of the glass product (200) by jetting pressurized airflow.

[0041] Reference Figure 2 An electrostatic generator (105) is located at the spraying port of the anti-mildew powder spreader (104). The electrostatic generator (105) contains multiple electrodes (105'), and each electrode (105') is insulated from the others.

[0042] A positioning sensor is used to acquire the position of the glass product (200). Specifically, the positioning sensor can be implemented using a photoelectric position sensor, a rotation angle sensor of a conveyor belt or roller, etc. In one embodiment, a photoelectric position sensor is used, in which case multiple photoelectric position sensors can be set at different positions on the glass product conveying mechanism (102), and the position of the glass product (200) is determined based on whether the photoelectric position sensor detects the glass product (200) and the position of the sensor itself. In another embodiment, a rotation angle sensor of a conveyor belt or roller is used to position the glass product (200), in which case the position of the glass product (200) can be determined from a designated starting position where the glass product (200) is placed based on the rotation angle of the conveyor belt or roller.

[0043] A charge-induction array sensor is used to detect charge distribution data on the surface of a glass product (200). Specifically, the charge-induction array sensor should be positioned before the dispensing port of the anti-mildew powder applicator (104).

[0044] The controller is also used to control the anti-mildew powder applicator (104) and the electrostatic generator (105), wherein the voltage of each electrode (105') of the electrostatic generator (105) is controlled according to the position and charge distribution data of the glass product (200) to realize the anti-mildew powder application method on the surface of the glass product as will be described in detail below.

[0045] In one embodiment, the anti-mildew powder applicator (100) for glass products also includes a depth sensor, which is used to collect depth data of the surface of the glass product (200). The controller also controls the voltage of each electrode of the electrostatic generator (105) based on the depth data. Specifically, the depth sensor can be a laser scanning sensor, a structured light sensor, a multi-view visible light camera, etc.

[0046] In one embodiment, the anti-mold powder spreading machine (100) for glass products further includes a laser light source and an image sensor. The laser light source is used to irradiate the anti-mold powder spread by the anti-mold powder spreader (104) onto the surface of the glass product (200). The image sensor is used to acquire images of the anti-mold powder spreading when the anti-mold powder spreader (104) spreads the anti-mold powder onto the surface of the glass product (200). The controller also controls the voltage of each electrode (105') of the electrostatic generator (105) according to the anti-mold powder spreading image to realize the anti-mold powder spreading method for glass products as described in detail below.

[0047] In one embodiment, there are multiple anti-mildew powder applicators (104) and electrostatic generators (105).

[0048] It is understood that the glass product surface anti-mildew powder spreading machine described above is only one possible implementation method. However, the method provided by the present invention is not limited to the glass product surface anti-mildew powder spreading machine described above. Other glass product surface anti-mildew powder spreading machines capable of implementing this method can also be used.

[0049] The steps of the method for applying anti-mildew powder to the surface of glass products provided by the present invention will be described below. (Refer to...) Figure 3 The method specifically includes:

[0050] Step S101: Obtain depth data and charge distribution data of the glass product surface.

[0051] Specifically, depth data of the glass surface can be collected using methods such as time-of-flight (ToF) measurement, interferometric laser phase measurement, structured light scanning, and stereo vision.

[0052] In one embodiment, the ToF measurement method is used. A laser scanning sensor can be selected as the depth sensor. The laser scanning sensor emits a laser to the surface of the glass product and measures the laser reflection time at each surface position. The depth data of the glass product surface is determined based on the laser reflection time at each surface position.

[0053] In one embodiment, an interferometric laser phase measurement method is employed. A laser scanning sensor can be selected as the depth sensor. The laser scanning sensor emits a laser beam onto the surface of the glass product and measures the laser reflection phase at each surface location. The depth data of the glass product surface is determined based on the laser reflection phase at each surface location. Specifically, a laser scanning sensor with a wavelength of 532 nanometers can be used. The phase difference of the reflected laser is calculated using a phase-shift interferometry algorithm. Combined with the calibrated optical path difference conversion formula Δh=(λ·Δφ) / 4π, the phase difference Δφ is converted into a height difference Δh.

[0054] In both embodiments described above, the depth sensor can be positioned at a fixed location on the glass product surface anti-mildew powder applicator. The glass product is moved by the glass product conveying mechanism of the applicator to collect depth data at all locations on the glass product surface. Alternatively, the glass product surface anti-mildew powder applicator can have a movable depth sensor, allowing for the collection of depth data at different locations on the glass product surface through sensor movement.

[0055] In one embodiment, a structured light scanning method is used. A structured light sensor can be selected as a depth sensor. The structured light sensor emits structured light onto the surface of the glass product and captures a structured light image of the glass product surface. The depth data of the glass product surface is determined based on the difference between the captured structured light image and the emitted structured light pattern.

[0056] In one embodiment, a stereo vision method is used. A multi-view visible light camera can be selected as the depth sensor. The multi-view visible light camera captures visible light images of the glass product surface at different positions. By comparing the differences between the visible light images captured simultaneously by the visible light cameras at different positions, the depth data of the product surface to be inspected is generated using triangulation.

[0057] Specifically, a charge induction array sensor can be used to scan the surface of a glass product, collect charge distribution data in analog signal form at each location, and then obtain charge density distribution data in digital signal form through analog-to-digital conversion (ADC).

[0058] Step S102: Calculate the curvature data of the glass product surface based on the depth data.

[0059] The process of determining the curvature data of a glass surface based on depth data can be specifically described as follows: The Delaunay triangulation algorithm can be used to construct a triangular mesh model from the discrete depth data (for example, constructing a triangular mesh model where the side length of each triangular facet is controlled within 15 micrometers). The local surface equation z = ax² + by² + cxy + dx + ey + f is fitted using the least squares method, and the Gaussian curvature at the vertices is calculated as K = (4ac - b²) / (1 + d² + e²)³. However, considering that the depth data acquired by the depth sensor may contain errors or erroneous data, which could lead to errors or erroneous curvature data, further analysis and processing of the curvature data are needed to eliminate potentially erroneous curvature data and perform reasonable corrections. The surface of the glass product can be further divided into multiple tiny curvature analysis regions. A quadratic differential operator is used to calculate the curvature gradient magnitude ‖∇K‖ = √((∂K / ∂x)² + (∂K / ∂y)² + (∂K / ∂z)²) within each curvature analysis region. When the curvature gradient magnitude ‖∇K‖ within a curvature analysis region exceeds a preset curvature gradient magnitude threshold, that region is marked as an abnormal curvature analysis region. The curvature data of the abnormal curvature analysis region is then replaced with the curvature data of adjacent curvature analysis regions (e.g., selecting the left-hand adjacent curvature analysis region) or with the average curvature data of all adjacent curvature analysis regions. Ultimately, curvature data in the form of a vector distribution map containing the coordinates of curvature extrema and gradient directions can be obtained. In the above data processing, the Compute Unified Device Architecture (CUDA) parallel computing architecture of NVIDIA's Graphics Processing Unit (GPU) can be used to accelerate depth data processing.

[0060] Step S103: Based on the curvature data, the surface of the glass product is divided into multiple meshes using finite element methods, and the target electric field intensity is determined for each mesh on the surface of the glass product.

[0061] Specifically, COMSOL multiphysics simulation software can be used to perform finite element mesh generation on the surface of the glass product. In COMSOL, a glass dielectric environment with a dielectric constant ε=5.8 is set, and the finite element mesh is automatically refined for raised regions with a curvature radius smaller than a preset curvature radius threshold. The process of determining the target electric field intensity for each mesh on the glass product surface can begin by selecting a mesh with 0 curvature and the lowest or highest depth as the reference mesh. The target electric field intensity of the reference mesh is set to a preset electric field intensity reference value. Then, starting from the reference mesh, the target electric field intensity of adjacent meshes is determined sequentially based on the target electric field intensity of the previous mesh and the curvature of the adjacent mesh, using a preset relationship between curvature and electric field intensity. Finally, the target electric field intensity corresponding to each mesh is obtained. To facilitate user monitoring of equipment operation, a target electric field intensity map data, including the target electric field intensity amplitude, equipotential lines, and gradient vectors, can be drawn based on the target electric field intensity corresponding to each mesh. This data is then rendered on a visualization interface using an OpenGL rendering engine for user viewing, achieving dynamic visualization.

[0062] Step S104: Determine the first voltage corresponding to each grid according to the target electric field intensity corresponding to each grid.

[0063] In one implementation, a first mapping relationship between electric field strength and a first voltage can be established through pre-calibration. Based on this first mapping relationship, the first voltage U1 applied by the electrostatic generator to each grid is determined according to the corresponding target electric field strength E. The first mapping relationship can be fitted to experimental data using the least squares method. For example, the first mapping relationship can be a quadratic function:

[0064]

[0065] in The first voltage corresponding to the grid. E The target electric field intensity corresponding to the grid. For reference electric field strength, , , All are preset voltage parameters, and the unit is V.

[0066] In one embodiment, the target electric field intensity corresponding to each grid is input into a pre-trained first fully connected neural network model to obtain the first voltage corresponding to each grid output by the first fully connected neural network model. The first fully connected neural network model is trained using pre-calibrated first training data. The first fully connected neural network model is trained using the target electric field intensity of the first training data as the input sample feature and the first voltage of the first training data as the output sample label.

[0067] Step S105: Determine the second voltage corresponding to each grid based on the charge distribution data.

[0068] In one implementation, a second mapping relationship between charge density and a second voltage can be established through pre-calibration. Based on this second mapping relationship, the corresponding second voltage U2 is determined for each grid cell according to the corresponding charge density distribution data ρ. The second mapping relationship can be fitted to experimental data using the least squares method. For example, the second mapping curve can be:

[0069]

[0070] in, The second voltage corresponding to the grid. ρ The charge density distribution data corresponding to the grid. For reference charge density, , All are preset voltage parameters, and the unit is V.

[0071] In one embodiment, the charge distribution data corresponding to each grid is input into a pre-trained second fully connected neural network model to obtain the second voltage corresponding to each grid output by the second fully connected neural network model. The second fully connected neural network model is trained using pre-labeled second training data. The model uses the charge distribution data from the second training data as input sample features and the second voltage from the second training data as the output sample label for training.

[0072] Specifically, the first fully connected neural network model can sequentially include an input layer, multiple levels of hidden layers, and an output layer. Since multi-level fully connected neural network models can fit nonlinear curves well, a fully connected neural network model with a large number of hidden layers can be set to fit the relationship between the target electric field strength and the first voltage. Each hidden layer further includes a fully connected layer and a batch normalization (BN) layer. Each fully connected layer contains multiple neurons, each representing a mapping of an activation function. Each neuron in the fully connected layer is connected to every neuron in the previous layer, automatically extracting the output features of the previous layer. The BN layer transforms the output values ​​of the fully connected layer into a standard normal distribution through normalization, preventing gradient explosion or vanishing, and also preventing overfitting, thus accelerating training convergence. (Reference) Figure 4 The training process of a fully connected neural network (corresponding to the first / second fully connected neural network mentioned above) can specifically include:

[0073] Step S201: Obtain pre-labeled training data to construct training samples and test samples. The first training data corresponding to the first fully connected neural network includes the target electric field strength and the first voltage; the sample feature is the target electric field strength, and the sample label is the first voltage. The second training data corresponding to the second fully connected neural network includes charge distribution data and the second voltage; the sample feature is the charge distribution data, and the sample label is the second voltage.

[0074] Specifically, after establishing the samples, the hold-out method can be used to divide the samples into training and testing sets according to a certain ratio (e.g., a 5:1 ratio of training to testing samples). Alternatively, cross-validation can be used to divide the samples into multiple sets. In each training round, one set is selected as the testing set, and the remaining sets form the training set. The testing set differs in different training rounds.

[0075] Step S202: Using the sample features of the training samples as the input of the fully connected neural network model and the sample labels of the training samples as the output of the fully connected neural network model as the objective, the fully connected neural network model is trained to obtain the training result.

[0076] Specifically, the training process of a fully connected neural network is a forward propagation process. The forward propagation process includes a linear bias process and a nonlinear fitting process of the neuron's activation function. The formula for the linear bias process is as follows:

[0077]

[0078] in, A The input vector for the fully connected layer has dimension . , n The number of samples; W Here is the weight matrix of the fully connected layer, with dimension 1. ; B The bias matrix has dimensions of . ; C This is the result of linear bias.

[0079] Activation functions can include Sigmoid, Rectified Linear Unit (ReLU), Leaky Rectified Linear Unit (Leaky ReLU), Hyperbolic Tangent Tanh, Softplus, and other functions.

[0080] After the fully connected layer extracts the output features of the previous layer, the Batch Normalization (BN) layer normalizes the output value of the corresponding fully connected layer using the following formula:

[0081]

[0082] In the above BN layer implementation formula, n The feature quantity input to the current batch of the BN layer. It is the first one entered in the current batch. i One characteristic, yes The corresponding BN layer output value, It is the average value of the features input in the current batch. It is the feature variance of the current batch of inputs. and These are the parameters of the trained fully connected neural network model. It is a very small preset parameter.

[0083] Step S203: Test the fully connected neural network model using the test sample set to obtain the model evaluation index.

[0084] Specifically, the following can be used: Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination. (R squared, Coefficient of determination) and other metrics are used as model evaluation indicators.

[0085] Step S204: Determine whether the training termination condition is met.

[0086] Specifically, the training termination condition can be that the model evaluation index is less than the corresponding preset termination threshold, that is, the model evaluation index represents whether the training termination condition is met; or the training termination condition can be set as the model evaluation index being less than the corresponding preset termination threshold or the number of training iterations reaching a preset threshold.

[0087] If the training termination condition is met, proceed to step S205; if the training termination condition is not met, proceed to step S206.

[0088] Step S205: Determine the loss function value corresponding to the training result, and update the model parameters of the fully connected neural network model based on the loss function value. Return to step S201 again.

[0089] Specifically, loss functions can include MAE, MSE, RMSE, MAPE, Huber loss, and Logarithm of the hyperbolic cosine (Log-Cosh) loss.

[0090] The Huber loss function is:

[0091]

[0092] The Huber loss function combines the advantages of MAE and MSE. It is effective when the distance between the model output value and the sample label is less than a preset parameter. In practice, MSE is used when the distance between the model output value and the sample label is greater than the preset parameter. In practice, MAE is used, which makes the Huber loss function more robust to training samples.

[0093] The Log-Cosh loss function is:

[0094]

[0095] The Log-Cosh loss function has similar advantages to the Huber loss function, and compared with the Huber loss function, it is quadratically differentiable across the entire range.

[0096] In the above loss function formula, i This is the sample number. For training samples i Sample characteristics, For fully connected neural network models based on The output value, For training samples i Sample labels, M Preset parameters.

[0097] Specifically, the model parameters of a fully connected neural network model can be updated using any of the following methods:

[0098] (1) Batch gradient descent algorithm:

[0099] Batch gradient descent calculates the loss function value based on the entire training sample set, calculates the gradient of the loss function value on each model parameter, and then updates the model parameters once.

[0100] Specifically, for any given model parameter, it is updated according to the following formula:

[0101]

[0102] in, For model parameters Version number, For the updated model parameters, These are the model parameters before the update. For learning rate, This is the gradient vector.

[0103] (2) Stochastic Gradient Descent Algorithm:

[0104] Because the batch gradient descent algorithm requires training all training samples once before updating the model parameters, the algorithm converges slowly when there are a large number of training samples.

[0105] Stochastic gradient descent calculates the loss function value by randomly selecting only one sample at a time and then updates the parameters once, making it faster than batch gradient descent.

[0106] Specifically, for any model parameter, the batch gradient descent algorithm used for updating the formula is the same, which will not be elaborated here.

[0107] (3) Mini-batch Gradient Descent Algorithm:

[0108] Mini-batch gradient descent is a compromise between batch gradient descent and stochastic gradient descent. It divides all training samples into multiple groups, calculates the loss function value after each group is trained, and then updates the parameters. Mini-batch gradient descent inherits the advantages of both the smooth descent of the loss curve in batch gradient descent and the high iterative efficiency of stochastic gradient descent.

[0109] Specifically, for any model parameter, the batch gradient descent algorithm used for updating the formula is the same, which will not be elaborated here.

[0110] (4) Momentum Gradient Descent Algorithm:

[0111] Unlike the previous three algorithms, the Momentum algorithm uses the concept of momentum, which makes the directional jitter less severe when the loss function value curve decreases.

[0112] Specifically, for any given model parameter, it is updated according to the following formula:

[0113]

[0114] in, For model parameters Version number, For the updated model parameters, These are the model parameters before the update. For learning rate, The gradient vector, For momentum hyperparameters, This is the amount of parameter update for this time. This represents the parameter update amount from the last time.

[0115] (5) Adaptive Learning Rate Optimization (AdaGrad) Algorithm:

[0116] Specifically, for any given model parameter, it is updated according to the following formula:

[0117]

[0118] in, For model parameters Version number, For the updated model parameters, These are the model parameters before the update. For learning rate, The gradient vector, This is the amount of parameter update for this time. This is the amount of the parameter update from the last time. As a preset constant, This is the element-wise multiplication of vectors.

[0119] For each iteration, use Vector records all previous gradients Accumulation of squared values ​​(second momentum), initial time The initial value is 0. Then, when calculating the update, the learning rate is calculated based on the accumulated squared gradient values ​​of each parameter. Parameters with larger accumulated gradient values ​​correspond to smaller learning rates, and vice versa. In the Momentum algorithm, all model parameters share the same learning rate, while the AdaGrad algorithm solves the problem of all parameters sharing the same learning rate.

[0120] (6) Root Mean Square Propagation (RMSProp) Algorithm:

[0121] Specifically, for any given model parameter, it is updated according to the following formula:

[0122]

[0123] in, For model parameters Version number, For the updated model parameters, These are the model parameters before the update. For learning rate, The gradient vector, For momentum hyperparameters, This is the amount of parameter update for this time. This is the amount of the parameter update from the last time. As a preset constant, This is the element-wise multiplication of vectors.

[0124] The RMSPro algorithm incorporates the concept of momentum into the gradient accumulation calculation, using hyperparameters. Control the weights of the current gradient value and the historically accumulated gradient values, accumulating only the previous ones each time. Use only a portion of the value, not the entire value, so as not to cause... The problem arises when the learning rate keeps decreasing due to continuous accumulation and increase, for example, if the gradient value of a certain parameter is small at this time. The accumulated value at the corresponding position will also decrease, instead of increasing indefinitely. The RMSProp algorithm solves the problem of the AdaGrad algorithm's learning rate continuously decreasing and slowing down in later stages.

[0125] (7) Adam algorithm:

[0126] Specifically, for any given model parameter, it is updated according to the following formula:

[0127]

[0128] in, For model parameters Version number, For the updated model parameters, These are the model parameters before the update. For learning rate, The gradient vector, , For momentum hyperparameters, This is the first-order update amount of the parameters. This is the first-order parameter update amount from the previous iteration. This is the second-order update amount of the parameters. This is the amount of the second-order parameter update from the previous iteration. As a preset constant, This is the element-wise multiplication of vectors.

[0129] The Adam algorithm adaptively adjusts the learning rate by calculating the first and second moment estimates of the gradient. The first moment estimate acts like momentum, smoothing gradient changes, while the second moment estimate adaptively adjusts the learning rate to suit different parameter characteristics. A correction step corrects for large estimation biases in the initial stage. The Adam algorithm combines the advantages of AdaGrad and RMSProp, considering both adaptive learning rate and momentum gradient, demonstrating good performance for training fully connected neural network models.

[0130] Step S206: End training of the fully connected neural network model.

[0131] Step S106: Calculate the difference between the first voltage and the second voltage for each grid cell to obtain the third voltage.

[0132] Specifically, for a given grid, the first voltage corresponding to the grid is the target voltage estimated based on the shape of the grid, while the second voltage corresponding to the grid is the equivalent voltage generated by the electrostatic generator estimated based on the electrostatic charge currently carried by the grid. Subtracting the second voltage U2 from the first voltage U1 gives the actual third voltage U3 that the electrostatic generator needs to apply to the anti-mold powder when it is sprinkled onto the grid, so as to achieve the desired amount of anti-mold powder being electrostatically adsorbed on the surface of the glass product.

[0133] Step S107: Apply anti-mildew powder to each area of ​​the glass product surface using the first application method.

[0134] The first application method involves the following: when the anti-mold powder applicator applies anti-mold powder to the target area on the surface of the glass product, the voltage of each electrode of the electrostatic generator is controlled to be the third voltage of the target grid corresponding to the electrode. This ensures that the anti-mold powder, carrying electrostatic charge, is adsorbed onto the target grid when applied through the corresponding electrode to the corresponding target grid. The target grid is the grid contained within the target area.

[0135] Specifically, the voltage amplitude of each electrode of the electrostatic generator can be controlled by a DAC (Digital-to-Analog Converter) to be the third voltage of the target grid corresponding to the electrode. If the glass product surface anti-mold powder application method is implemented using the glass product surface anti-mold powder application machine provided in the first aspect, then during the movement of the glass product via the glass product conveying mechanism, the glass product sequentially passes through the application port of the anti-mold powder applicator from the top position to the bottom position. Correspondingly, each grid on the surface of the glass product will sequentially pass through the application port of the anti-mold powder applicator. Based on the current position of the glass product, the target area to which the anti-mold powder is applied at the current application port position can be determined, and thus, the target grids to which the anti-mold powder is applied at the current application port position can be determined. Correspondingly, each electrode of the electrostatic generator corresponds to a target grid within the target area. At this time, by controlling the voltage of each electrode of the electrostatic generator to be the third voltage of the target grid corresponding to the current electrode, the electrostatic charge carried by the anti-mold powder in each target grid corresponding to the current anti-mold powder application position can be precisely controlled to be the desired amount of charge. Thus, the anti-mold powder is adsorbed on the corresponding target grid with the desired amount of electrostatic charge, achieving the effect of uniform application of anti-mold powder. It will not cause uneven application of anti-mold powder in different positions due to the uneven shape of the glass product surface and the actual electrostatic charge it carries.

[0136] During the process of applying anti-mildew powder to the surface of glass products through the above steps, uneven application may still occur due to factors such as errors in sensor measurement data and differences between the actual voltage applied by the electrostatic generator and the expected third voltage when applying the anti-mildew powder. Therefore, in one embodiment, after completing the above steps, the method may further include the following steps:

[0137] Step S108: Apply anti-mildew powder to each area of ​​the glass product surface using the second application method.

[0138] The second method of dissemination is:

[0139] Based on images of the anti-mold powder being spread by the powder spreader captured by the image sensor, the actual amount of anti-mold powder adsorbed in each grid is determined by particle tracking velocimetry (PTV). The fourth voltage corresponding to each grid is determined based on the amount of anti-mold powder adsorbed in each grid. When the anti-mold powder spreader spreads anti-mold powder onto the target area on the surface of the glass product, the voltage of each electrode of the electrostatic generator is controlled to be the fourth voltage of the target grid corresponding to the electrode, so that when the anti-mold powder is spread to the corresponding target grid through the corresponding electrode, it carries electrostatic charge and is adsorbed on the target grid.

[0140] Specifically, based on images of the anti-mold powder being dispensed by the powder dispenser captured by an image sensor, the actual amount of anti-mold powder adsorbed in each grid can be monitored using particle tracking velocimetry (PTV), thus determining the differences in actual adsorption amounts between different grids. A laser light source can be used to illuminate the area between the powder dispenser and the glass surface. Then, a high-speed image sensor captures images of the powder being dispensed at extremely short time intervals (e.g., time intervals corresponding to a frame rate of 1000 fps). PTV technology is used to analyze each frame to calculate the velocity of each powder particle at the image capture location. Based on the velocity of each powder particle, a trajectory is fitted, and the landing point of each powder particle is finally determined. By combining the landing points of each powder particle with the average mass of the anti-mold powder, the total amount of anti-mold powder ultimately adsorbed in each grid area can be calculated. Considering that both the anti-mold powder and the glass surface carry a certain amount of electrostatic charge, and that the glass surface may also have a certain amount of electrostatic charge, there is an electromagnetic force between the anti-mold powder and the glass surface that affects the trajectory of each powder. To simplify the influence of these interaction forces on the trajectory of the anti-mold powder, the high-speed image sensor needs to be set at a position where the image position of the captured horizontal plane is not high above the glass surface (for example, when spreading anti-mold powder on the target area of ​​the glass surface, the distance between the height of the horizontal plane captured by the image sensor and the average or highest height of the target area is less than a preset height difference threshold, such as setting the preset height difference threshold to 1 cm). After calculating the velocity at the distance from the captured horizontal plane using PTV technology, the trajectory between the captured horizontal plane position and the final landing position of the anti-mold powder is simplified and estimated as a straight line, a parabola, or other simpler dynamic trajectory. Based on the velocity of the anti-mold powder, the captured position, and the simplified estimated dynamic trajectory, the landing position of the anti-mold powder is finally estimated. Then, based on the landing position of each anti-mold powder, the amount of anti-mold powder adsorbed in each grid on the glass surface is determined. For specific implementation methods of PTV technology, please refer to existing technologies; this article will not elaborate further.

[0141] Generally, the amount of anti-mold powder adsorbed across different grids follows a normal distribution. Therefore, grids with an actual amount of anti-mold powder adsorbed greater than or equal to μ-3σ (where μ is the mean of the amount of anti-mold powder adsorbed across all grids and σ is the standard deviation of the amount of anti-mold powder adsorbed across all grids) can be identified as normal grids, while grids with an actual amount of anti-mold powder adsorbed less than μ-3σ can be identified as abnormal grids. For abnormal grids, due to insufficient actual adsorption of anti-mold powder, more anti-mold powder needs to be applied during the second application to compensate for the difference with normal grids. This requires applying more electrostatic charge to the applied anti-mold powder during the second application to increase the amount of anti-mold powder adsorbed by the abnormal grids. For normal grids, although a relatively uniform and expected amount of anti-mold powder has been adsorbed during the first application, abnormal grids require further application during the second application. Since the anti-mold powder applicator for glass products cannot apply anti-mold powder only to abnormal grids and not to normal grids, the second application necessitates reapplying anti-mold powder to normal grids. Therefore, during the second application, a small but uniform amount of anti-mold powder can be applied to the normal grids to ensure uniform adsorption of anti-mold powder between normal and abnormal grids after the second application. This can be achieved by pre-calibrating and establishing an anti-mold powder adsorption difference. The third mapping relationship with the fourth voltage U4 is used to determine the corresponding fourth voltage U4 for each grid based on the corresponding anti-mildew powder adsorption difference. Specifically, for a given grid, the anti-mildew powder adsorption difference... It is the difference between the average actual amount of anti-mildew powder adsorbed, μ, and the actual amount of anti-mildew powder adsorbed, G, in this grid. The third mapping relationship is a piecewise function curve: The third mapping relationship is:

[0142]

[0143]

[0144] in, The fourth voltage corresponding to a certain grid. μ The value represents the average amount of anti-mildew powder adsorbed across all grids. σ The standard deviation of the amount of anti-mildew powder adsorbed across all grids. G represents the adsorption difference of the anti-mildew powder corresponding to the grid, and G is the actual amount of anti-mildew powder adsorbed by the grid. The reference voltage indicates the voltage at which the adsorption amount equals the mean, i.e. The initial voltage setting value at 0; , It is a dimensionless proportionality coefficient. This indicates the sensitivity of the adsorption amount deviation to voltage adjustment. This indicates the sensitivity of voltage adjustment when there is severe insufficient adsorption; , This is the voltage offset used to compensate for system errors, measured in volts (V). Typically... To avoid overshoot, for example 0.433.

[0145] In one embodiment, since the above steps involve applying anti-mold powder to the glass product surface a second time after the first application, the process is more efficient. Figure 1 The illustrated anti-mildew powder applicator for glass products can be equipped with only one set of anti-mildew powder applicator (104) and electrostatic generator (105). The glass product conveying mechanism (102) moves the glass product (200) in the opposite direction to the first anti-mildew powder applicator process so that the glass product (200) moves again to the applicator (104) applicator position for a second applicator of anti-mildew powder.

[0146] In one embodiment, such as using Figure 1 The illustrated anti-mildew powder applicator for glass products can be equipped with at least two sets of anti-mildew powder applicators (104) and electrostatic generators (105). The glass product (200) is moved from the applicator opening of the first anti-mildew powder applicator (104) to the applicator opening of the second anti-mildew powder applicator (104) by the glass product conveying mechanism (102) so that the second anti-mildew powder applicator (104) can apply anti-mildew powder to the glass product (200) for the second time.

[0147] Reference Figure 5 This invention provides a controller for applying anti-mildew powder to the surface of glass products, comprising:

[0148] Measurement module M1 is used to acquire depth data and charge distribution data on the surface of glass products;

[0149] Curvature calculation module M2 is used to calculate the curvature data of the glass product surface based on the depth data;

[0150] The finite element calculation module M3 is used to divide the surface of the glass product into multiple meshes based on the curvature data, and to calculate the target electric field intensity of each mesh.

[0151] The first voltage calculation module M4 is used to determine the first voltage corresponding to each grid according to the target electric field intensity corresponding to each grid.

[0152] The second voltage calculation module M5 is used to determine the second voltage corresponding to each grid based on the charge distribution data;

[0153] The third voltage calculation module M6 is used to calculate the difference between the corresponding first voltage and the corresponding second voltage for each grid to obtain the third voltage;

[0154] The first powder-spraying module M7 is used to spray anti-mildew powder onto various areas of the surface of the glass product in a first spraying method. The first spraying method is as follows: when the anti-mildew powder sprayer sprays anti-mildew powder onto the target area of ​​the surface of the glass product, it controls the voltage of each electrode of the electrostatic generator to be the third voltage of the target grid corresponding to the electrode, so that when the anti-mildew powder is sprayed to the corresponding target grid through the corresponding electrode, it carries electrostatic charge and is adsorbed on the target grid. The target grid is the grid contained in the target area.

[0155] The electrostatic generator is located at the spraying port of the anti-mildew powder applicator, and the electrodes of the electrostatic generator are mutually insulated.

[0156] In one embodiment, the controller further includes:

[0157] The second powder-spreading module M8 is used to spread anti-mildew powder to various areas of the glass product surface in a second spreading method. The second spreading method is as follows: based on the image of the anti-mildew powder spreader spreading anti-mildew powder acquired by the image sensor, the actual adsorption amount of anti-mildew powder in each grid is determined by particle tracking velocity measurement (PTV); based on the adsorption amount of anti-mildew powder in each grid, the fourth voltage corresponding to each grid is determined; when spreading anti-mildew powder to the target area on the surface of the glass product, the voltage of each electrode of the electrostatic generator is controlled to be the fourth voltage of the target grid currently corresponding to the electrode, so that when the anti-mildew powder is spread to the corresponding target grid through the corresponding electrode, it carries electrostatic charge and is adsorbed on the target grid.

[0158] It should be noted that the controller embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the controller embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0159] It should be noted that the anti-mildew powder application controller for glass products provided in this embodiment of the invention is used to execute all the process steps of the anti-mildew powder application method for glass products described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0160] Fourthly, the present invention also provides an electronic device. This electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data acquisition program. When the processor executes the computer program, it implements the steps in the above-described embodiments of the method for spraying anti-mildew powder onto the surface of glass products, for example... Figure 3 The step S101 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the measurement module M101.

[0161] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0162] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0163] The processor can be a Central Processing Unit (CPU), or 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. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0164] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0165] If the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0166] The present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the method for applying anti-mildew powder to the surface of glass products as described in the second aspect.

[0167] Based on this understanding, the present invention can implement all or part of the processes in the methods of the above embodiments, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0168] The present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being used to implement the method of applying anti-mildew powder to the surface of glass products as described in the second aspect when the computer is run.

[0169] In this invention, when spreading anti-mildew powder onto the surface of glass products, the voltage of the electrodes of the electrostatic generator that applies electrostatic charge to the anti-mildew powder is adaptively adjusted according to the shape of the glass product surface and the electrostatic charge carried by the glass product. This allows the anti-mildew powder spread at different locations on the surface of the glass product to be adsorbed based on the corresponding electrostatic charge, resulting in uniform spreading of the anti-mildew powder on the surface of the glass product.

[0170] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for applying anti-mildew powder to the surface of glass products, characterized in that, include: Acquire depth and charge distribution data on the surface of glass products; The curvature data of the glass product surface is calculated based on the depth data; The surface of the glass product is divided into multiple meshes using finite element methods based on the curvature data, and the target electric field intensity of each mesh is calculated. The first voltage corresponding to each grid is determined based on the target electric field intensity corresponding to each grid. The second voltage corresponding to each grid is determined based on the charge distribution data; For each grid cell, the difference between the first voltage and the second voltage is calculated to obtain the third voltage. Anti-mildew powder is spread to various areas of the surface of the glass product using a first spreading method, wherein the first spreading method is as follows: when the anti-mildew powder spreader spreads anti-mildew powder to the target area on the surface of the glass product, the voltage of each electrode of the electrostatic generator is controlled to be the third voltage of the target grid corresponding to the electrode, so that when the anti-mildew powder is spread to the corresponding target grid through the corresponding electrode, it carries electrostatic charge and is adsorbed on the target grid, wherein the target grid is the grid contained in the target area; The electrostatic generator is located at the spraying port of the anti-mildew powder spreader, and the electrodes of the electrostatic generator are mutually insulated.

2. The method according to claim 1, characterized in that, The step of determining the first voltage corresponding to each grid based on the target electric field intensity corresponding to each grid includes: Based on a preset first mapping relationship, the first voltage corresponding to each grid is determined for the target electric field intensity corresponding to each grid. The first mapping relationship is as follows: The first voltage corresponding to a certain grid. E The target electric field intensity corresponding to the grid. For reference electric field strength, , , All are preset voltage parameters.

3. The method according to claim 1, characterized in that, The step of determining the first voltage corresponding to each grid based on the target electric field intensity corresponding to each grid includes: The target electric field intensity corresponding to each grid is input into the pre-trained first fully connected neural network model to obtain the first voltage corresponding to each grid output by the first fully connected neural network model. The first fully connected neural network model is trained using the target electric field intensity of the first training data as the input sample feature and the first voltage of the first training data as the output sample label.

4. The method according to claim 1, characterized in that, The step of determining the second voltage corresponding to each grid based on the charge distribution data includes: Based on a preset second mapping relationship, the second voltage corresponding to each grid is determined according to the charge distribution data corresponding to each grid. The second mapping relationship is as follows: in, The second voltage corresponding to a certain grid. ρ The charge density distribution data corresponding to the grid. For reference charge density, , All are preset voltage parameters.

5. The method according to claim 1, characterized in that, After applying the anti-mildew powder to various areas of the glass article surface using the first application method, the method further includes: Anti-mold powder is spread onto various areas of the glass product surface using a second spreading method. This second spreading method involves: determining the actual amount of anti-mold powder adsorbed by each grid using particle tracking velocity measurement (PTV) based on images of the anti-mold powder being spread by the powder spreader acquired by an image sensor; determining a fourth voltage corresponding to each grid based on the amount of anti-mold powder adsorbed by each grid; and controlling the voltage of each electrode of the electrostatic generator to be the fourth voltage of the target grid corresponding to the electrode when spreading the anti-mold powder to the corresponding target grid via the corresponding electrode, so that the anti-mold powder carries electrostatic charge and adsorbs onto the target grid when spread through the corresponding electrode.

6. The method according to claim 5, characterized in that, The step of determining the fourth voltage corresponding to each grid based on the amount of anti-mildew powder adsorbed in each grid includes: Based on the third mapping relationship, the corresponding fourth voltage is determined for each grid according to the corresponding amount of anti-mildew powder adsorption. The third mapping relationship is as follows: in, The fourth voltage corresponding to a certain grid. μ The value represents the average amount of anti-mildew powder adsorbed across all grids. σ The standard deviation of the amount of anti-mildew powder adsorbed across all grids. G represents the adsorption difference of the anti-mildew powder corresponding to the grid, and G is the actual amount of anti-mildew powder adsorbed by the grid. The reference voltage indicates the voltage at which the adsorption amount equals the mean, i.e. The initial voltage setting value at 0; , It is a dimensionless proportionality coefficient. This indicates the sensitivity of the adsorption amount deviation to voltage adjustment. This indicates the sensitivity of voltage adjustment when there is severe insufficient adsorption; , This is the voltage offset, used to compensate for system errors.

7. A controller for applying anti-mildew powder to the surface of glass products, characterized in that, include: The measurement module is used to acquire depth data and charge distribution data on the surface of glass products; The curvature calculation module is used to calculate the curvature data of the glass product surface based on the depth data. The finite element calculation module is used to divide the surface of the glass product into multiple meshes based on the curvature data, and to calculate the target electric field intensity of each mesh. The first voltage calculation module is used to determine the first voltage corresponding to each grid according to the target electric field intensity corresponding to each grid. The second voltage calculation module is used to determine the second voltage corresponding to each grid based on the charge distribution data; The third voltage calculation module is used to calculate the difference between the corresponding first voltage and the corresponding second voltage for each grid, so as to obtain the third voltage; The first powder-spreading module is used to spread anti-mildew powder to various areas of the surface of the glass product in a first spreading method. The first spreading method is as follows: when the anti-mildew powder spreader spreads anti-mildew powder to the target area on the surface of the glass product, it controls the voltage of each electrode of the electrostatic generator to be the third voltage of the target grid corresponding to the electrode, so that when the anti-mildew powder is spread to the corresponding target grid through the corresponding electrode, it carries electrostatic charge and is adsorbed on the target grid. The target grid is the grid contained in the target area. The electrostatic generator is located at the spraying port of the anti-mildew powder spreader, and the electrodes of the electrostatic generator are mutually insulated.

8. An electronic device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the method for applying anti-mildew powder to the surface of glass products as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the method for applying anti-mildew powder to the surface of glass products as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that is used to implement the method of applying anti-mildew powder to the surface of glass products as described in any one of claims 1-6 when the computer is run.

Images