Data Processing Method, Device, Equipment, Storage Medium and Product for Glue Dispensing

Abstract

The present invention discloses a data processing method, device, equipment, storage medium and product for dispensing, relating to the technical field of dispensing. The method includes: obtaining expected glue weight data and a preset prediction model, determining prediction glue type category data corresponding to the candidate dispensing parameter data based on the candidate dispensing parameter data and a preset glue type classification model in the preset prediction model, where the glue type categories include qualified and unqualified, determining prediction glue weight data corresponding to the candidate dispensing parameter data based on the candidate dispensing parameter data and a preset glue weight regression model in the preset prediction model, and determining target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data based on a first objective function determined according to the expected glue weight data, the prediction glue type category data and the prediction glue weight data. By adopting the above technical solution, the determination efficiency and accuracy of dispensing process parameters can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of dispensing, and particularly to a data processing method, device, equipment, storage medium and product for dispensing. Background Art

[0002] The dispensing process is an industrial manufacturing technology that realizes functions such as bonding, sealing, insulation or heat conduction by precisely controlling the application, potting or dropping of glue or other fluids, and can be widely applied to multiple fields and products.

[0003] The current dispensing control method relies on manual experience. According to the demand information of dispensing, manually combine and debug process parameters such as the heating temperature of the glue, the dispensing height, the valve opening time of the glue valve, the rising time, the delay time, the falling time and the plunger stroke. Since there are many types of the above parameters, and each parameter is not independent of the dispensing effect, it will affect the shape of the glue line width and height, etc., and a one-to-one linear function relationship cannot be formed. Therefore, it is very time-consuming and inefficient to debug qualified dispensing process parameters. Summary of the Invention

[0004] The present invention provides a data processing method, device, equipment, storage medium and product for dispensing, which can realize automatic determination of accurate dispensing parameters.

[0005] According to one aspect of the present invention, there is provided a data processing method for dispensing, including:

[0006] Obtain expected glue weight data and a preset prediction model, wherein the preset prediction model includes a preset glue type classification model and a preset glue weight regression model;

[0007] Based on the candidate dispensing parameter data and the preset glue type classification model, determine the predicted glue type category data corresponding to the candidate dispensing parameter data, wherein the glue type category includes qualified and unqualified;

[0008] Based on the candidate dispensing parameter data and the preset glue weight regression model, determine the predicted glue weight data corresponding to the candidate dispensing parameter data;

[0009] Based on a first objective function, determine the target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data, wherein the first objective function is determined according to the expected glue weight data, the predicted glue type category data and the predicted glue weight data.

[0010] According to another aspect of the present invention, there is provided a data processing device for dispensing. The device includes:

[0011] An acquisition module, configured to acquire expected glue weight data and a preset prediction model, wherein the preset prediction model includes a preset glue type classification model and a preset glue weight regression model;

[0012] A category prediction module, configured to determine predicted glue type category data corresponding to the candidate dispensing parameter data based on the candidate dispensing parameter data and the preset glue type classification model, wherein the glue type categories include qualified and unqualified;

[0013] A glue weight prediction module, configured to determine predicted glue weight data corresponding to the candidate dispensing parameter data based on the candidate dispensing parameter data and the preset glue weight regression model;

[0014] A parameter data determination module, configured to determine target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data based on a first objective function, wherein the first objective function is determined according to the expected glue weight data, the predicted glue type category data, and the predicted glue weight data.

[0015] According to another aspect of the present invention, there is provided an electronic device, the electronic device comprising:

[0016] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the data processing method for dispensing according to any embodiment of the present invention.

[0017] According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to execute the data processing method for dispensing according to any embodiment of the present invention when executed.

[0018] According to another aspect of the present invention, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the data processing method for dispensing according to any embodiment of the present invention.

[0019] In the technical solution of the embodiment of the present invention, expected glue weight data and a preset prediction model are obtained. Based on the candidate dispensing parameter data and the preset glue type classification model in the preset prediction model, the predicted glue type category data corresponding to the candidate dispensing parameter data is determined. The glue type categories include qualified and unqualified. Based on the candidate dispensing parameter data and the preset glue weight regression model in the preset prediction model, the predicted glue weight data corresponding to the candidate dispensing parameter data is determined. Based on the first objective function determined according to the expected glue weight data, the predicted glue type category data, and the predicted glue weight data, the target dispensing parameter data corresponding to the expected glue weight data is determined from the candidate dispensing parameter data. By adopting the above technical solution, the model is used to predict the candidate dispensing process parameters in terms of both glue type and glue weight, and the target function is used to automatically determine the target dispensing process parameters that meet the constraints of both glue type and glue weight, improving the determination efficiency and accuracy of the dispensing process parameters, reducing the process parameter debugging cost, improving the debugging efficiency, and further improving the production and manufacturing efficiency related to the dispensing process.

[0020] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0022] Figure 1 is a flowchart of a data processing method for dispensing according to an embodiment of the present invention;

[0023] Figure 2 is a schematic diagram of the generation process of a glue weight regression model according to an embodiment of the present invention;

[0024] Figure 3 is a schematic diagram of a model update process according to an embodiment of the present invention;

[0025] Figure 4 is a flowchart of a data processing method for dispensing according to an embodiment of the present invention;

[0026] Figure 5 is a schematic framework diagram of a data processing solution for dispensing according to an embodiment of the present invention;

[0027] Figure 6It is a schematic structural diagram of a data processing device for dispensing according to an embodiment of the present invention;

[0028] Figure 7 It is a schematic structural diagram of an electronic device for implementing the data processing method for dispensing according to an embodiment of the present invention. Detailed implementation manners

[0029] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0030] It should be noted that the terms "first", "second", "initial", and "target" in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0031] Figure 1 It is a flowchart of a data processing method for dispensing according to an embodiment of the present invention. This embodiment is applicable to the situation of automatically determining dispensing process parameters. This method can be executed by a data processing device for dispensing. The data processing device for dispensing can be implemented in the form of hardware and / or software. The data processing device for dispensing can be configured in an electronic device, and the electronic device can be a computer device or an industrial production device, etc. As Figure 1 shown, the method includes:

[0032] Step 101, obtain expected glue weight data and a preset prediction model, where the preset prediction model includes a preset glue type classification model and a preset glue weight regression model.

[0033] The glue weight generally refers to the actual weight of the glue or other fluids (collectively referred to as glue for ease of description) on the product during the dispensing process. It is an important parameter to ensure consistent glue volume for each dispensing, thus guaranteeing product quality. In the embodiments of the present invention, the expected glue weight data can be understood as the glue weight data that needs to be achieved after dispensing determined according to actual production requirements, and can also be called the target glue weight data, which can specifically be a weight value.

[0034] In the embodiments of the present invention, the preset prediction model can be a machine learning model obtained through pre-training, which can be trained based on the collected sample data. The sample data can include sample dispensing parameter data, sample glue type data, and sample glue weight data. The preset prediction model can be a dynamically updated model to improve prediction accuracy. The preset prediction model includes a preset glue type classification model and a preset glue weight regression model, which can be used to predict the glue type and glue weight respectively. The glue type categories can include qualified and unqualified. Optionally, the preset glue type classification model can include a classification model based on Extreme Gradient Boosting (XGBoost), or a logistic regression model for classification tasks; the preset glue weight regression model can include a regression model based on XGBoost, or a linear regression model for regression tasks.

[0035] Step 102: Based on the candidate dispensing parameter data and the preset glue type classification model, determine the predicted glue type category data corresponding to the candidate dispensing parameter data, where the glue type categories include qualified and unqualified.

[0036] Exemplarily, the dispensing parameter data can be the parameter values of the dispensing process parameters. The candidate dispensing parameter data can include the parameter values of any dispensing process parameters supported by the dispensing process equipment or the parameter values of the preset dispensing process parameters. The specific types of the involved dispensing process parameters are not limited, and can include glue-related control parameters or spray valve-related control parameters, etc. For example, it can include at least one of the glue heating temperature, dispensing height, valve opening time of the glue valve, rising time, delay time, falling time, and plunger stroke. Generally, there are multiple groups of candidate dispensing parameter data, and each group of dispensing parameter data usually includes the parameter values of multiple dispensing process parameters. A preset search algorithm (such as a heuristic search algorithm) can be used to sequentially determine the candidate dispensing parameter data currently participating in the calculation until the candidate dispensing parameter data that meets the requirements corresponding to the expected glue weight data is found, which is used as the dispensing parameter data obtained after prediction using the preset prediction model, and can also be denoted as the target dispensing parameter data.

[0037] Exemplarily, the glue type classification can be regarded as a binary classification task for predicting whether the glue type of the dispensed glue is qualified under specified dispensing process parameters. The preset glue type classification model can be trained using a training set.

[0038] Optionally, the preset glue type classification model includes a first glue type classification model, which is trained using a local training set, and specifically can be a classification model based on XGBoost.

[0039] Exemplarily, a sampling set of dispensing process parameters is preset, and data on glue types and glue weights under different dispensing parameter data are collected through dispensing experiments, and data preprocessing is performed to obtain a local data set , also known as the local training set, can be expressed as . Among them, represents the feature vector of the th sample (the feature dimension is , that is, the dimension of adjustable dispensing process parameters), R represents the set of real numbers, that is, the set of all real numbers; represents the glue weight value corresponding to the th sample; , represents the glue type label corresponding to the th sample, represents unqualified, represents qualified; represents the total number of samples in the local training set.

[0040] Exemplarily, the collected data is preprocessed, converted into binary classification labels (qualified and unqualified), and then trained using a machine learning model (such as XGBoost) to obtain a first glue type classification model.

[0041] Exemplarily, in the collection of glue type data, for each set of dispensing parameters, data of the first quantity (denoted as m, for example, m = 3) of glue lines can be sampled. Each glue line records the number of scatter points, the number of bubbles, and the glue width values at the second quantity (denoted as n, for example, n = 40) locations. For each set of parameter values recorded, those that meet the preset qualified conditions are recorded as qualified (category ), otherwise they are unqualified (category ). The preset qualified conditions can include conditions in multiple dimensions. If the conditions in multiple dimensions are simultaneously met, it is considered to meet the preset qualified conditions. The multiple dimensions can include, for example, the broken glue dimension, the bubble dimension, the scatter point dimension, and the glue width dimension. For the broken glue dimension, all the glue width values of the m groups of glue lines are not 0; for the bubble dimension, the average number of bubbles in the m groups of glue lines is less than the bubble threshold (such as 3); for the scatter point dimension, the average number of scatter points in the m groups of glue lines is less than the scatter point threshold (such as 3); for the glue width dimension, at least p groups (for example, 2 groups) of the m groups of glue lines are glue width uniform. Glue width uniformity means that after deleting the maximum and minimum values of the n glue width values, the standard deviation of the remaining glue width values divided by the average value of the remaining glue width values is less than the first ratio value (such as 5%), and the range divided by the average value is less than the second ratio value (such as 10%).

[0042] The first glue type classification model is a binary classification model , which can be expressed as . , is the probability that the glue type predicted by the classification model is qualified, is the optimal parameter of the model. The model output is the probability that the sample belongs to the category : , where represents the predicted probability that the th sample is a qualified glue type, represents the predicted probability that the th sample is an unqualified glue type, is the parameter of the classification model. The glue type category predicted by the model is .

[0043] The training objective of the first glue type classification model is to optimize the parameters of the classification model by minimizing the binary cross-entropy loss function:

[0044] .

[0045] where represents the true glue type label of the th sample; represents the predicted probability of the th sample.

[0046] XGBoost is a machine learning algorithm based on Gradient Boosting. It improves the prediction accuracy of the model by constructing and combining multiple weak classifiers (usually decision trees). The working principle of XGBoost for binary classification can be simply summarized in the following steps:

[0047] (1) Initialize the model: XGBoost starts with a simple model, such as all samples having the same predicted value. For binary classification problems, it is usually initialized by predicting all samples as a constant value (for example, the initial probability of all samples is 0.5).

[0048] (2) Calculate the error (residual): XGBoost calculates the error of the current prediction of the model. For binary classification problems, the loss function used is the log-loss function, and this error represents the degree of error when the current model predicts positive and negative classes.

[0049] (3) Build a new decision tree: Based on the error of the current model, XGBoost builds a new decision tree to correct this error. Each node of the decision tree represents a split on a certain feature, and the leaf stores the predicted correction value of the residual (that is, the direction and magnitude that the model needs to adjust).

[0050] (4) Update the model: The newly constructed decision tree fine-tunes the predicted value of each sample. The output of each new tree is an adjustment to the previous model, and XGBoost adds the new tree to the existing model, making the prediction results gradually approach a more accurate direction.

[0051] (5) Repeat the iteration: XGBoost repeats the above steps to gradually build new trees, and each iteration corrects based on the error information of the previous time. This process usually continues for multiple rounds until the model reaches the predetermined number of trees or the error drops to a certain threshold.

[0052] (6) Output the final prediction: For a binary classification problem, the final output is the probability that a sample belongs to a certain class. XGBoost sums up the prediction results of all trees with weights, and then through the logistic function (LogisticFunction, ), converts the result into a probability value between 0 and 1. Set a classification threshold, which is usually set to 0.5. If the probability is greater than the classification threshold, then predict as class, otherwise predict as class (i.e., unqualified).

[0053] Exemplarily, in this step, the candidate dispensing parameter data that needs to participate in the calculation currently is input into a preset glue type classification model, and the predicted glue type category data corresponding to the currently input candidate dispensing parameter data is determined according to the data output by the preset glue type classification model. The predicted glue type category data can be, for example, a probability value, and the specific glue type category can be determined according to the comparison result between the probability value and the preset threshold. For example, if the probability value is greater than or equal to 0.5, it means qualified, and if the probability value is less than 0.5, it means unqualified.

[0054] Step 103: Based on the candidate dispensing parameter data and the preset glue weight regression model, determine the predicted glue weight data corresponding to the candidate dispensing parameter data.

[0055] Exemplarily, glue weight prediction can be regarded as a regression task for predicting the glue weight of dispensing under specified dispensing process parameters, and the preset glue weight regression model can be trained using a training set.

[0056] Optionally, the preset glue weight regression model includes a first glue weight regression model, and the first glue weight regression model is trained using a local training set, specifically a regression model based on XGBoost.

[0057] Exemplarily, the collected data is preprocessed. For example, in the collection of glue weight data, each set of dispensing parameters is weighed a third number of times (denoted as q, e.g., q = 5), and the mean value of each weighing is recorded. The mean value of the q weighing values is denoted as the glue weight value of this set of dispensing parameters. Optionally, to ensure data quality, only the dispensing parameter groups with the standard deviation divided by the mean of the q weighings less than the third proportional value (such as 5%) may be retained.

[0058] Exemplarily, the first glue weight regression model is a function that maps from the feature space to the target space R: , represents the predicted glue weight data, is the optimal parameter of the model. For the th sample, the predicted glue weight value is: , where are the parameters of the model. The training objective of the regression model is to obtain the optimal parameters by minimizing the loss function of the glue weight prediction error (such as the mean squared error):

[0059] .

[0060] The principle of XGBoost in the regression task is similar to that in the classification task. It is still based on the idea of gradient boosting trees (GBDT), and reduces the model error by gradually constructing multiple weak regression models (decision trees). Its core idea is to correct the prediction error of the previous round in each round of iteration, and finally construct a powerful regression model. The working principle of XGBoost regression can be summarized into the following steps:

[0061] (1) Initialize the model: XGBoost starts from a simple model. Usually, the predicted values of all samples are the same, such as using the mean value of the target variable as the initial prediction.

[0062] (2) Calculate the error (residual): XGBoost calculates the current prediction error of the model (i.e., the difference between the predicted value and the true value), which is also called the residual. XGBoost uses the mean squared error (MSE) as the loss function for the regression task, that is, it minimizes the sum of the squared differences between the predicted value and the true value. These errors are used to measure the current prediction effect of the model.

[0063] (3) Build a new decision tree: XGBoost builds a new decision tree to reduce the current error. Each node of this tree represents a division of the features, and the value stored in each leaf node is the adjusted value of the residual (i.e., the amount that the model needs to correct).

[0064] (4) Update the model: The output of the new tree (the adjusted value of the residual) is added to the predicted value of the existing model to update the prediction result of the overall model.

[0065] (5) Repeated iteration: XGBoost continues to repeat the above steps, gradually constructing new decision trees. Each new tree is constructed based on the residuals of the current model. The output of each tree is used to further adjust the predicted values and correct the residuals.

[0066] (6) Output the final prediction: After multiple rounds of iteration, the prediction results of all trees (the output values of each tree) are weighted and summed to obtain the final regression prediction value.

[0067] Exemplarily, in this step, the candidate dispensing parameter data that needs to participate in the calculation currently is input into a preset glue weight regression model, and the predicted glue weight data corresponding to the currently input candidate dispensing parameter data is determined according to the data output by the preset glue weight regression model. The predicted glue weight data can be, for example, a glue weight value.

[0068] Step 104: Determine the target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data based on the first objective function, where the first objective function is determined according to the expected glue weight data, the predicted glue type category data, and the predicted glue weight data.

[0069] Exemplarily, the search for dispensing process parameters can be regarded as an optimization problem of high-dimensional continuous variables. Through a preset search algorithm, the optimal (minimizing the first objective function) dispensing parameter combination (target dispensing parameter data) is effectively found in the high-dimensional continuous parameter space (which can be regarded as a data space containing all candidate dispensing parameter data). The optimization goal is to find a set of dispensing parameter data such that the preset glue type classification model predicts it as qualified and the specification (predicted glue weight) predicted by the preset glue weight regression model is closest to the target specification (expected glue weight), which can be expressed as a multi-objective optimization problem:

[0070] .

[0071] Wherein, is the glue weight value predicted by the preset glue weight regression model, that is, the predicted glue weight data, is the target glue weight value, that is, the expected glue weight data, is the regression error, such as the mean square error or the squared error, such as , is the output of the preset glue type classification model (1 represents qualified, 0 represents unqualified), is the penalty factor for the glue type qualified item ( ), () is the indicator function.

[0072] Exemplarily, start a heuristic search algorithm based on a machine learning prediction model, and randomly sample a set of process parameters As the initial point (i.e., the first determined current candidate dispensing parameter data), after iterative optimization, a set of optimal process parameters is obtained. , where:

[0073] is the optimization objective function (the first objective function) targeted at the penalty factor for the qualified item of the glue type ( ), () is the indicator function, .

[0074] Exemplarily, considering the high dimensionality and complexity of the parameter space and avoiding global traversal, a heuristic algorithm (such as Bayesian optimization) can be used for parameter search. Of course, other search algorithms, such as grid search or random search, etc., can also be adopted. Bayesian optimization is an efficient method for optimizing complex functions. Its core idea is to use a surrogate model (such as Gaussian process or random forest, etc.) to approximate the objective function and continuously update this model through Bayesian to find the optimal solution.

[0075] The search process of Bayesian optimization can be summarized as follows: First, by sampling the initial points of the objective function, a surrogate model is trained using these points, which can give the predicted mean and uncertainty of the objective function at different inputs. Then, using an acquisition function (such as expected improvement or upper confidence bound, etc.), the next point to be evaluated is selected. This step balances exploring new regions and exploiting known good results based on the predicted values and uncertainties of the surrogate model in the unexplored regions. Evaluate the new point and update the surrogate model. Continuously repeat this process until the optimal value of the objective function is found.

[0076] The data processing method for dispensing provided by the embodiments of the present invention obtains expected glue weight data and a preset prediction model. Based on the candidate dispensing parameter data and the preset glue type classification model in the preset prediction model, the predicted glue type category data corresponding to the candidate dispensing parameter data is determined. The glue type category includes qualified and unqualified. Based on the candidate dispensing parameter data and the preset glue weight regression model in the preset prediction model, the predicted glue weight data corresponding to the candidate dispensing parameter data is determined. Based on the first objective function determined according to the expected glue weight data, the predicted glue type category data, and the predicted glue weight data, the target dispensing parameter data corresponding to the expected glue weight data is determined from the candidate dispensing parameter data. By adopting the above technical solutions, the model is used to predict the candidate dispensing process parameters in terms of both glue type and glue weight, and the objective function is used to automatically determine the target dispensing process parameters that meet the constraints of both glue type and glue weight, improving the determination efficiency and accuracy of the dispensing process parameters, reducing the process parameter debugging cost, improving the debugging efficiency, and further improving the production and manufacturing efficiency related to the dispensing process.

[0077] Over time, the equipment will experience wear and tear during the production process. With the same dispensing parameters, there will be significant differences in the dispensing effect after a certain period (such as one month). At this time, it is necessary to readjust the process parameters. Whenever there is a new dispensing requirement, the existing solutions need to re-debug the valve process parameters on-site, which affects the project progress.

[0078] In the embodiments of the present invention, considering that due to uncontrollable factors such as machine wear, environmental changes (such as temperature changes), and glue characteristics (such as inconsistent viscosities), the distribution of the collected glue weight data will shift over time. To reduce the impact of this shift on the model, a deviation correction model is introduced to improve the prediction accuracy of the glue weight regression model in the latest environment. Specifically, a basic model can be constructed through a large amount of stable data collected in the early stage, and then the deviation regression model is trained with the latest offset data. The two are combined to construct a correction model to improve the generalization ability and prediction accuracy of the model on the latest data.

[0079] In some embodiments, the first glue weight regression model is trained in the following manner: the local training set is divided into a first training set and a second training set, where the sampling time of any sample data in the second training set is later than the sampling time of any sample data in the first training set; the first glue weight regression sub-model is trained using the first training set; the second training set is predicted using the first glue weight regression sub-model to obtain predicted values, and the prediction errors are calculated based on the predicted values; the second glue weight regression sub-model is trained using the prediction errors and the second training set; the first glue weight regression model is determined based on the first glue weight regression sub-model and the second glue weight regression sub-model.

[0080] Exemplarily, is divided into (the first training set, equivalent to the early data) and (the second training set, equivalent to the later data, where the glue weight distribution has shifted). The specific division method is not limited. For example, the sample data in are sorted according to the sampling time of each sample data, and a split point is found. The first training set is determined based on the sample data before the split point, and the second training set is determined based on the sample data after the split point. First, a basic model is trained, such as training a basic regression model with strong generality based on (the first glue weight regression sub-model) to capture the overall relationship between the dispensing parameters and the glue weight. Then, the deviation of the later data is calculated, such as predicting using to obtain the predicted value Exemplarily, The sample dispensing parameter data in the In The output of is used as the predicted value and the exist The prediction error on , that is, calculation The difference between the sample weight value and the predicted value is used to obtain the prediction error. Then the deviation prediction model is trained based on The prediction error on training the bias regression model for later data (Second regressor model), the second regressor model can specifically be a regression model based on XGBoost. The training process can refer to the relevant description in the previous article.

[0081] Exemplarily, the first glue weight regression model is determined according to the first glue weight regression sub-model and the second glue weight regression sub-model, and specifically, the first glue weight regression model is determined according to the sum of the first glue weight regression sub-model and the second glue weight regression sub-model. , that is, .

[0082] One of the keys to the bias correction regression model is the separation of the basic data set and the bias data set. In the embodiment of the present invention, an optimal change point position can be found by detecting the change point of the regression error sequence, and the sequence can be divided into two parts so that the distribution of the two parts is as different as possible, thereby achieving data separation.

[0083] In some embodiments, the local training set is divided into a first training set and a second training set, including: using an initial regression model to determine the regression error of each sample data in the local training set; sorting the regression errors according to the sampling time of each sample data in the local training set to obtain a regression error sequence; determining a target regression error in the regression error sequence, and dividing the local training set into a first training set and a second training set according to the target regression error, wherein the sample data corresponding to the target regression error is the last sample data in the first training set or the first sample data in the second training set, and the target regression error is determined with the goal of maximizing the difference between the sample distribution of the first training set and the sample distribution of the second training set. In this way, the training set can be segmented efficiently and accurately, and the model prediction accuracy can be improved.

[0084] The initial glue weight regression model may be a glue weight regression model to be updated, for example, it may be a first glue weight regression model obtained by first training with a local training set or a first glue weight regression model that needs to be updated in other cases.

[0085] Figure 2 is a schematic diagram of a process of generating a regression model according to an embodiment of the present invention.Figure 2 As shown, the regression error (such as the squared error) of each sample data in the local training set (original data) is calculated using the initial glue weight regression model. The regression errors are sorted according to the sampling time of each sample data in the local training set to obtain a regression error sequence. The regression error sequence is used for error single change point detection, that is, to determine the target regression error. The target regression error is used for data classification, and the local training set is divided into early-stage data (the first training set) and late-stage data (the second training set). The basic regression model (the first glue weight regression sub-model) is trained using the early-stage data. The prediction deviation is calculated using the basic regression model and the late-stage data, and the deviation regression model (the second glue weight regression sub-model) is trained using the prediction deviation. Finally, the deviation correction regression model (the first glue weight regression model) is obtained.

[0086] Exemplarily, the regression error sequence can be denoted as . The error single change point detection process includes: finding a change point position , maximizing the segmentation loss function , where is the data fitting loss function, such as the within-segment mean squared error, , , where i, j, and k represent positive integers. Subsequently, data separation is performed. According to the optimal change point position , is divided into the early-stage basic data set (the first training set) and the late-stage bias data set (the second training set) .

[0087] To further control the impact of the prediction deviation on the target search, the embodiments of the present invention can correct the model in a timely manner while ensuring the search efficiency. Therefore, a method of online updating by mixing strong and weak models is adopted to achieve model correction.

[0088] In some embodiments, the preset glue type classification model is determined by the weighted sum of the first glue type classification model and the second glue type classification model, and the preset glue weight regression model is determined by the weighted sum of the first glue weight regression model and the second glue weight regression model; the first glue type classification model and the first glue weight regression model are trained using the local training set, the second glue type classification model and the second glue weight regression model are trained using the online training set, and the online training set includes the test result data obtained after performing glue dispensing tests using at least one of the target glue dispensing parameter data; the training cost of the second glue type classification model is lower than the training cost of the first glue type classification model, and the training cost of the second glue weight regression model is lower than the training cost of the first glue weight regression model. Thus, on the basis of timely correcting the model and ensuring the search efficiency, the impact of the prediction deviation on the search for the target glue dispensing parameter data can be further controlled, and the prediction accuracy can be improved.

[0089] Among them, the online training set can be a dynamically updated training set. After each prediction using a preset prediction model, target dispensing parameter data can be obtained. By using the target dispensing parameter data for dispensing tests, corresponding test result data can be obtained. Based on the test result data, online training samples are constructed and stored in the online training set to dynamically update the online training set. For example, the online data set is denoted as , for the searched target dispensing process parameters , perform dispensing tests in real time to obtain test glue type data (glue type result ) and test glue weight data (glue weight value ), and record the test result data , and update the test result data to the online database .

[0090] Exemplarily, the training cost can be measured based on model complexity and training cost, etc. Those with high training costs can be denoted as strong models, including the first glue weight regression model (denoted as ) and the first glue type classification model (denoted as ); those with low training costs can be denoted as weak models, including the second glue weight regression model (denoted as ) and the second glue type classification model (denoted as ). Among them, the strong model selects a model with high model complexity and high prediction accuracy, such as the classification model based on XGBoost, the regression model based on XGBoost, and the bias correction regression model based on XGBoost mentioned above; the weak model selects a model with low model complexity and low training cost, such as the logistic regression model for classification tasks and the linear regression model for regression tasks. Based on the data set , perform weak model training to obtain the weak classification model and the weak regression model , which are used to capture the simple relationship between machine parameters and the latest glue weight and glue type. Optionally, in the case where the target dispensing parameter data fails the dispensing test, the weak model is updated using the current online training set. By mixing strong and weak models to correct the prediction bias. At the same time, the real-time update of the weak model ensures the real-time correction of the dispensing test results to the algorithm, and the performance evaluation and automatic update of the strong model can ensure the accuracy of the prediction model. Since the training time cost of the weak model is low, and the automatic update of the strong model can be performed only when the model evaluation fails, the training frequency is low. Therefore, the online hybrid correction model update mechanism realizes the real-time improvement of the model accuracy at an extremely low time cost.

[0091] Exemplarily, the prediction results of the strong model and the weak model are weighted and summed to correct the model results.

[0092] For the classification model, , ;

[0093] For the regression model, .

[0094] Among them, and are the importance weights of the strong model and the weak model, , , when , the weak model has no effect on the result. It should be noted that for the classification model and the regression model, can take different values.

[0095] In some embodiments, it may further include: in response to the performance evaluation result of the first glue type classification model and / or the first glue weight regression model failing, moving the online training set into the local training set to obtain a new local training set; training the first glue type classification model and / or the first glue weight regression model based on the new local training set to obtain an updated first glue type classification model and / or an updated first glue weight regression model. Thus, the local training set used to train the strong model can be updated in a timely manner, supplementing the sample data in the more timely online training set to the local training set, and improving the prediction accuracy of the updated strong model.

[0096] Exemplarily, a performance evaluation trigger condition for the first glue type classification model and / or the first glue weight regression model can be preset, denoted as the preset evaluation trigger condition, and the performance evaluation for the first glue type classification model and / or the first glue weight regression model is triggered when the preset evaluation trigger condition is met, and the specific evaluation method is not limited. Among them, the preset evaluation trigger condition can be set based on the time condition and / or the number of times the target dispensing parameter data fails the corresponding dispensing test. For example, the time interval between the current time and the last performance evaluation time is greater than the preset interval threshold, such as 1 month, or the continuous cumulative number of times the target dispensing parameter data fails the corresponding dispensing test is greater than the preset number threshold, such as 3 times.

[0097] Exemplarily, the online training set can be used to evaluate the performance of the first glue type classification model and / or the first glue weight regression model. For example, for the first glue type classification model, if the precision of the model is greater than the precision threshold, it is considered passed, otherwise, it is considered failed; for the first glue weight regression model, if the mean absolute error (MAE) of the model is less than the mean absolute error threshold, it is considered passed, otherwise, it is considered failed. Among them, , .

[0098] Exemplarily, if the strong model fails the performance evaluation, the online database is merged into the local database. , , where represents the empty set; and based on train a strong model to obtain a new strong classification model and / or a new strong regression model , that is, obtain an updated first glue type classification model and / or an updated first glue weight regression model.

[0099] Figure 3 is a schematic diagram of a model update process provided by an embodiment of the present invention. As Figure 3 shown, update the online database according to the dispensing test results, that is, update the online training set, and use the online training set to train a weak model. For example, when the target dispensing parameter data fails the dispensing test, update the weak model using the current online training set. Evaluate the performance of the strong model. If it passes, there is no need to update the strong model. If it fails, update the local database according to the online database and the local database, that is, update the local training set, and use the updated local data set to train the strong model. The weak model and the strong model form a hybrid correction model, that is, a preset prediction model.

[0100] Figure 4 is a flowchart of a data processing method for dispensing provided by an embodiment of the present invention. This embodiment is optimized based on the above optional embodiments and adds relevant processing steps for the case where the dispensing test fails. Figure 5 is a schematic framework diagram of a data processing solution for dispensing provided by an embodiment of the present invention, which can be combined with Figure 5 to understand the embodiments of the present invention.

[0101] As Figure 4 shown, the method includes:

[0102] Step 401, obtain expected glue weight data and a preset prediction model, where the preset prediction model includes a preset glue type classification model and a preset glue weight regression model.

[0103] Exemplarily, as Figure 5 shown, in the offline startup phase, collect data for model training to obtain a machine learning prediction model, that is, a preset prediction model. In the online application phase, input the target glue weight, that is, input the expected glue weight data.

[0104] Step 402, based on the candidate dispensing parameter data and the preset glue type classification model, determine the predicted glue type category data corresponding to the candidate dispensing parameter data.

[0105] Step 403, based on the candidate dispensing parameter data and the preset glue weight regression model, determine the predicted glue weight data corresponding to the candidate dispensing parameter data.

[0106] Step 404: Determine the target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data, where the first objective function is determined based on the expected glue weight data, the predicted glue type category data, and the predicted glue weight data.

[0107] As Figure 5 shown, according to the input target glue weight, using a search algorithm combined with a machine learning prediction model, the optimal process parameters, that is, the target dispensing parameter data, are searched for.

[0108] Step 405: Obtain the test result data after performing a dispensing test using the target dispensing parameter data.

[0109] Exemplarily, the target dispensing parameter data is denoted as , and based on a dispensing test is performed to obtain a glue type result and a glue weight value .

[0110] Step 406: Add the test result data to the online training set.

[0111] Exemplarily, update the online training set: .

[0112] Step 407: Determine whether the target dispensing parameter data meets the preset requirements based on the difference between the test glue weight data and the expected glue weight data in the test result data, as well as the test glue type data in the test result data.

[0113] As Figure 5 shown, after obtaining the optimal process parameters, it is judged whether the target is met, that is, it is determined whether the target dispensing parameter data meets the preset requirements.

[0114] Exemplarily, evaluate the test result data. If the glue type is qualified ( ) and the glue weight error is less than the preset error threshold, it is determined that the target dispensing parameter data meets the preset requirements, that is, the target is met; otherwise, it is determined that the target dispensing parameter data does not meet the preset requirements, that is, the target is not met. Among them, the preset error threshold can be set according to actual needs, such as 5%, .

[0115] Step 408: If it does not meet the requirements, train the preset prediction model based on the online training set to obtain an updated preset prediction model.

[0116] As Figure 5 shown, if it meets the preset requirements, that is, the target is met, the optimal process parameters If it does not meet the preset requirements, that is, it does not achieve the goal, the preset prediction model is trained based on the online training set to obtain an updated preset prediction model. Specifically, it can be to train and update the second glue type classification model and the second glue weight regression model in the preset prediction model based on the online training set. Exemplarily, denote the preset glue type classification model and the preset glue weight regression model included in the preset prediction model as and respectively. Based on the online training set train to obtain and , and let , .

[0117] Step 409: Select target sample data from the online training set based on the second objective function, where the second objective function is determined according to the expected glue weight data, the sample glue weight values of the sample data in the online training set, and the sample glue types.

[0118] Exemplarily, after the preset prediction model is updated, the new target dispensing parameter data corresponding to the expected glue weight data can be determined again based on the updated preset prediction model. In order to further improve the search efficiency, a more accurate search starting point can be determined so as to search for the new target dispensing parameter data more quickly.

[0119] Exemplarily, select from the online training set the sample dispensing parameter data that minimizes the second objective function (target sample data), where () is an indicator function.

[0120] Step 410: Use the dispensing parameter data in the target sample data as the search starting point, and based on the updated preset prediction model again, determine the new target dispensing parameter data corresponding to the expected glue weight data.

[0121] Exemplarily, let , that is, use the dispensing parameter data in the target sample data as the search starting point, and based on the updated preset prediction model again, determine the new target dispensing parameter data corresponding to the expected glue weight data until the dispensing test passes, that is, obtain the target dispensing parameter data that meets the preset requirements.

[0122] The data processing method for dispensing provided by the embodiments of the present invention uses a strong and weak hybrid model that supports dynamic update to predict candidate dispensing process parameters in terms of both glue type and glue weight, and uses a search algorithm to automatically determine the target dispensing process parameters that meet the constraints of both glue type and glue weight by using an objective function. After determining the target dispensing process parameters, a dispensing test is carried out, and the dispensing test results are detected. If they do not meet the preset requirements, the preset prediction model can be updated through an online training set, and a more accurate search starting point can be selected from the online training set. Among them, the glue weight prediction model in the strong model includes a deviation regression part. If changes occur in the glue state, environmental state, equipment state, etc., the model algorithm can automatically correct the hyperparameters of the model to adapt to the latest changes, quickly meet new dispensing requirements, further improve the determination efficiency and accuracy of dispensing process parameters, reduce the debugging cost of process parameters, improve the debugging efficiency, and further improve the production and manufacturing efficiency related to the dispensing process.

[0123] Figure 6 is a schematic structural diagram of a data processing device for dispensing provided by the embodiments of the present invention. As Figure 6 shown, the device includes:

[0124] An acquisition module 601, configured to acquire expected glue weight data and a preset prediction model, where the preset prediction model includes a preset glue type classification model and a preset glue weight regression model;

[0125] A category prediction module 602, configured to determine predicted glue type category data corresponding to the candidate dispensing parameter data based on the candidate dispensing parameter data and the preset glue type classification model, where the glue type categories include qualified and unqualified;

[0126] A glue weight prediction module 603, configured to determine predicted glue weight data corresponding to the candidate dispensing parameter data based on the candidate dispensing parameter data and the preset glue weight regression model;

[0127] A parameter data determination module 604, configured to determine target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data based on a first objective function, where the first objective function is determined according to the expected glue weight data, the predicted glue type category data, and the predicted glue weight data.

[0128] The data processing device for dispensing according to the embodiments of the present invention uses a model to predict candidate dispensing process parameters in terms of both glue type and glue weight, and automatically determines the target dispensing process parameters that meet the constraints of both glue type and glue weight by using an objective function, improves the determination efficiency and accuracy of dispensing process parameters, reduces the debugging cost of process parameters, improves the debugging efficiency, and thus can improve the production and manufacturing efficiency related to the dispensing process.

[0129] Optionally, the preset glue weight regression model includes a first glue weight regression model, which is trained as follows: divide the local training set into a first training set and a second training set, where the sampling time of any sample data in the second training set is later than that of any sample data in the first training set; use the first training set to train a first glue weight regression sub-model; use the first glue weight regression sub-model to predict the second training set to obtain predicted values, and calculate prediction errors based on the predicted values; use the prediction errors and the second training set to train a second glue weight regression sub-model; determine the first glue weight regression model according to the first glue weight regression sub-model and the second glue weight regression sub-model.

[0130] Optionally, the step of dividing the local training set into a first training set and a second training set includes: using the initial glue weight regression model to determine the regression errors of the sample data in the local training set; sorting the regression errors according to the sampling time of the sample data in the local training set to obtain a regression error sequence; determining the target regression error in the regression error sequence, and dividing the local training set into a first training set and a second training set according to the target regression error, where the sample data corresponding to the target regression error is the last sample data of the first training set or the first sample data of the second training set, and the target regression error is determined with the goal of maximizing the difference between the sample distributions of the first training set and the second training set.

[0131] Optionally, the preset glue type classification model is determined by the weighted sum of a first glue type classification model and a second glue type classification model, and the preset glue weight regression model is determined by the weighted sum of a first glue weight regression model and a second glue weight regression model; the first glue type classification model and the first glue weight regression model are trained using the local training set, and the second glue type classification model and the second glue weight regression model are trained using an online training set, where the online training set includes test result data obtained after performing glue dispensing tests using at least one of the target dispensing parameter data; the training cost of the second glue type classification model is lower than that of the first glue type classification model, and the training cost of the second glue weight regression model is lower than that of the first glue weight regression model.

[0132] Optionally, the device further includes:

[0133] A test result data acquisition module, configured to acquire test result data obtained after performing a glue dispensing test using the target dispensing parameter data, after determining the target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data based on the first objective function.

[0134] An online training set adding module, configured to add the test result data to an online training set;

[0135] A preset requirement determining module, configured to determine whether the target dispensing parameter data meets the preset requirements according to the difference between the test glue weight data and the expected glue weight data in the test result data, and the test glue type data in the test result data;

[0136] A model updating module, configured to train the preset prediction model based on the online training set to obtain an updated preset prediction model when it is determined that the target dispensing parameter data does not meet the preset requirements;

[0137] A target sample data obtaining module, configured to select target sample data from the online training set based on a second objective function, where the second objective function is determined according to the expected glue weight data, the sample glue weight values of the sample data in the online training set, and the sample glue type;

[0138] A re - determining module, configured to re - determine the new target dispensing parameter data corresponding to the expected glue weight data based on the updated preset prediction model with the dispensing parameter data in the target sample data as the search starting point.

[0139] Optionally, the apparatus further includes:

[0140] A local training set updating module, configured to move the online training set into a local training set to obtain a new local training set in response to the performance evaluation result of the first glue type classification model and / or the first glue weight regression model failing to pass;

[0141] A training module, configured to train the first glue type classification model and / or the first glue weight regression model based on the new local training set to obtain an updated first glue type classification model and / or an updated first glue weight regression model.

[0142] The data processing apparatus for dispensing provided by the embodiments of the present invention can execute the data processing method for dispensing provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

[0143] Figure 7FIG. 0 shows a schematic structural diagram of an electronic device 10 that can be used to implement embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0144] As Figure 7 shown, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0145] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0146] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the data processing method for dispensing glue.

[0147] In some embodiments, the data processing method for dispensing can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the data processing method for dispensing described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the data processing method for dispensing by any other suitable means (e.g., by means of firmware).

[0148] The various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), systems on a chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0149] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs can be executed entirely on the machine, partially on the machine, as a stand-alone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0150] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0151] In order to provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0152] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0153] A computing system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The relationship between the client and the server is created by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0154] Embodiments of the present disclosure provide a computer program product including a computer program which, when executed by a processor, implements the data processing method for dispensing provided in the above embodiments.

[0155] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is imposed herein.

[0156] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A data processing method for dispensing glue, characterized in that Including: Obtain expected glue weight data and a preset prediction model, where the preset prediction model includes a preset glue type classification model and a preset glue weight regression model, and the expected glue weight data is the data of the glue weight that needs to be achieved after dispensing; Based on the candidate dispensing parameter data and the preset glue type classification model, determine the predicted glue type category data corresponding to the candidate dispensing parameter data; Based on the candidate dispensing parameter data and the preset glue weight regression model, determine the predicted glue weight data corresponding to the candidate dispensing parameter data; Based on a first objective function, determine the target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data, where the first objective function is determined according to the expected glue weight data, the predicted glue type category data, and the predicted glue weight data; Wherein, the preset glue weight regression model includes a first glue weight regression model, and the first glue weight regression model is trained through the following method: Divide the local training set into a first training set and a second training set, where the sampling time of any sample data in the second training set is later than the sampling time of any sample data in the first training set; Use the first training set to train a first glue weight regression sub-model; Use the first glue weight regression sub-model to predict the second training set to obtain predicted values, and calculate prediction errors according to the predicted values; Use the prediction errors and the second training set to train a second glue weight regression sub-model; Determine the first glue weight regression model according to the first glue weight regression sub-model and the second glue weight regression sub-model.

2. The data processing method for dispensing according to claim 1, wherein The dividing the local training set into a first training set and a second training set includes: Use an initial glue weight regression model to determine the regression errors of each sample data in the local training set; Sort the regression errors according to the sampling time of each sample data in the local training set to obtain a regression error sequence; Determine the target regression error in the regression error sequence, and divide the local training set into a first training set and a second training set according to the target regression error, where the sample data corresponding to the target regression error is the last sample data of the first training set or the first sample data of the second training set, and the target regression error is determined with the goal of maximizing the difference between the sample distributions of the first training set and the second training set.

3. The dispensing data processing method according to claim 1, characterized in that, The preset glue type classification model is determined by the weighted sum of a first glue type classification model and a second glue type classification model, and the preset glue weight regression model is determined by the weighted sum of a first glue weight regression model and a second glue weight regression model; the first glue type classification model and the first glue weight regression model are trained using the local training set, and the second glue type classification model and the second glue weight regression model are trained using an online training set, and the online training set includes test result data obtained after dispensing tests using at least one of the target dispensing parameter data; the training cost of the second glue type classification model is lower than the training cost of the first glue type classification model, and the training cost of the second glue weight regression model is lower than the training cost of the first glue weight regression model.

4. The dispensing data processing method according to claim 1, characterized in that, After determining the target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data based on the first objective function, the following steps are further included: Obtain the test result data after performing a dispensing test using the target dispensing parameter data; Add the test result data to the online training set; Determine whether the target dispensing parameter data meets the preset requirements according to the difference between the test glue weight data in the test result data and the expected glue weight data, and the test glue type data in the test result data; If not, train the preset prediction model based on the online training set to obtain an updated preset prediction model; Select target sample data from the online training set based on a second objective function, where the second objective function is determined according to the expected glue weight data, the sample glue weight values of the sample data in the online training set, and the sample glue type; Taking the dispensing parameter data in the target sample data as the search starting point, re-determine the new target dispensing parameter data corresponding to the expected glue weight data based on the updated preset prediction model.

5. The dispensing data processing method according to claim 3, wherein The following steps are also included: In response to the performance evaluation result of the first glue type classification model and / or the first glue weight regression model not passing, move the online training set into the local training set to obtain a new local training set; Train the first glue type classification model and / or the first glue weight regression model based on the new local training set to obtain an updated first glue type classification model and / or an updated first glue weight regression model.

6. A data processing device for dispensing glue, characterized in that, It includes: An acquisition module, configured to acquire expected glue weight data and a preset prediction model, where the preset prediction model includes a preset glue type classification model and a preset glue weight regression model, and the expected glue weight data is the data of the glue weight that needs to be achieved after dispensing; A category prediction module, configured to determine the predicted glue type category data corresponding to the candidate dispensing parameter data based on the candidate dispensing parameter data and the preset glue type classification model; A glue weight prediction module, configured to determine the predicted glue weight data corresponding to the candidate dispensing parameter data based on the candidate dispensing parameter data and the preset glue weight regression model; A parameter data determination module, configured to determine the target dispensing parameter data corresponding to the expected glue weight data from the candidate dispensing parameter data based on a first objective function, where the first objective function is determined according to the expected glue weight data, the predicted glue type category data, and the predicted glue weight data; Among them, the preset glue weight regression model includes a first glue weight regression model, and the first glue weight regression model is trained through the following method: Divide the local training set into a first training set and a second training set, where the sampling time of any sample data in the second training set is later than the sampling time of any sample data in the first training set; Train a first glue weight regression sub-model using the first training set; Use the first glue weight regression sub-model to predict the second training set to obtain predicted values, and calculate prediction errors according to the predicted values; Train a second glue weight regression sub-model using the prediction errors and the second training set; Determine a first glue weight regression model according to the first glue weight regression sub-model and the second glue weight regression sub-model.

7. An electronic device, characterized in that, The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the dispensing data processing method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions for causing a processor to implement the dispensing data processing method according to any one of claims 1-5 when executed.

9. A computer program product, characterized in that, The computer program product includes a computer program which, when executed by a processor, implements the dispensing data processing method according to any one of claims 1-5.

Images