Product defect detection method, device, equipment and medium

Abstract

The invention relates to a product defect detection method, device and equipment and a medium. According to the method, the equipment parameter set which is generated in the product production process in the preset time period and consists of the data corresponding to the multiple equipment parameters is acquired, and the data corresponding to each equipment parameter is arranged according to the time sequence, so that the product detection accuracy is improved based on the equipment parameter set through the pre-trained product detection model; and determining a detection label for indicating that the product is good or bad, and outputting a product detection result of the product based on the detection label of the product. According to the invention, the method can achieve the prediction of the good or bad products through a machine learning method, can timely find out the products which are likely to be bad, and can timely carry out the early warning of the products which are likely to be bad.

Description

Technical Field

[0001] The present invention relates to the field of display technologies, and in particular, to a method, device, equipment, and medium for detecting product defects. Background Art

[0002] During the production process of products, product defects may occur due to various factors such as material quality, process control, and equipment failures. Therefore, there is an urgent need for a method for detecting product defects to timely identify products that may have defects, so as to timely issue warnings for products that may have defects. Summary of the Invention

[0003] The present invention provides a method, device, equipment, and medium for detecting product defects to solve the deficiencies in the related art.

[0004] According to a first aspect of an embodiment of the present invention, a method for detecting product defects is provided. The method includes:

[0005] Obtain a set of equipment parameters during the product production process within a preset time period. The set of equipment parameters includes data corresponding to multiple equipment parameters. For any equipment parameter, the data corresponding to the equipment parameter is arranged in time sequence;

[0006] Based on the set of equipment parameters, determine a detection label of the product through a pre-trained product detection model. The detection label is used to indicate whether the product is good or defective;

[0007] Based on the detection label of the product, output a product detection result of the product.

[0008] According to a second aspect of an embodiment of the present invention, a device for detecting product defects is provided. The device includes:

[0009] An obtaining module, configured to obtain a set of equipment parameters during the product production process within a preset time period. The set of equipment parameters includes data corresponding to multiple equipment parameters. For any equipment parameter, the data corresponding to the equipment parameter is arranged in time sequence;

[0010] A determining module, configured to determine a detection label of the product through a pre-trained product detection model based on the set of equipment parameters. The detection label is used to indicate whether the product is good or defective;

[0011] An output module, configured to output a product detection result of the product based on the detection label of the product.

[0012] According to a third aspect of an embodiment of the present invention, a computing device is provided. The computing device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the operations performed by the method for detecting product defects provided in the first aspect are implemented.

[0013] According to a fourth aspect of an embodiment of the present invention, there is provided a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, the operations performed by the product defect detection method provided in the first aspect as described above are implemented.

[0014] According to the above embodiments, the present invention obtains a set of device parameters composed of data corresponding to multiple device parameters generated during the production process of a product within a preset time period, and the data corresponding to each device parameter is arranged in time sequence. Then, based on the set of device parameters, through a pre-trained product detection model, a detection label for indicating whether the product is good or defective is determined. Furthermore, based on the detection label of the product, a product detection result of the product is output. The present invention can predict whether a product is good or defective through a machine learning method to timely discover products that may be defective, so as to timely give a warning to products that may be defective.

[0015] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention.

[0017] Figure 1 is a flowchart of a product defect detection method shown according to an embodiment of the present invention.

[0018] Figure 2 is a schematic diagram of an interface of a second screening interface shown according to an embodiment of the present invention.

[0019] Figure 3 is a schematic diagram of an interface of a detection data selection interface shown according to an embodiment of the present invention.

[0020] Figure 4 is a schematic diagram of a sample distribution diagram shown according to an embodiment of the present invention.

[0021] Figure 5 is a schematic diagram of an interface of a third screening interface shown according to an embodiment of the present invention.

[0022] Figure 6 is a flowchart of a model training process shown according to an embodiment of the present invention.

[0023] Figure 7 is a block diagram of a product defect detection device shown according to an embodiment of the present invention.

[0024] Figure 8 It is a schematic structural diagram of a computing device shown according to an embodiment of the present invention. Detailed implementation manners

[0025] Here, exemplary embodiments will be described in detail, and examples thereof are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0026] The present invention provides a method for detecting product defects, which can predict product defects during production through machine learning, so as to timely discover products that may have defects, and thus can timely give early warnings to products that may have defects, effectively reducing the occurrence of product defects.

[0027] Optionally, the product defect detection method provided by the present invention can be used to inspect organic light-emitting diode (OLED) products (such as OLED glass panels) during production to timely discover OLED products that may have defects.

[0028] The above product defect detection method can be executed by a computing device, and the computing device can be a terminal device, such as a desktop computer, a portable computer, a smart phone, a tablet computer, etc. Optionally, the computing device can also be a server, such as a single server, multiple servers, a server cluster, etc. The present invention does not limit the device type and the number of devices of the computing device.

[0029] See Figure 1 , Figure 1 is a flowchart of a product defect detection method shown according to an embodiment of the present invention. As Figure 1 shown, the method includes:

[0030] Step 101, obtain a set of device parameters during the product production process within a preset time period. The set of device parameters includes data corresponding to multiple device parameters. For any device parameter, the data corresponding to the device parameter is arranged in time sequence.

[0031] Optionally, the preset time period can be a time period corresponding to a first duration before the current moment, and the first duration can be any duration.

[0032] Among them, the set of device parameters may include data corresponding to multiple device parameters in the product production process, and the present invention does not limit the specific type of device parameters. It should be noted that the device parameters in the product production process may be control target value (Set Value, SV) parameters. Taking the product as an OLED product as an example, the device parameters in the product production process may include temperature (such as heater temperature, cooler temperature, etc.), flow rate (such as gas flow rate, vacuum pump pumping rate, etc.), pressure (such as atmosphere pressure, vacuum degree, etc.), but not limited thereto.

[0033] It should be noted that the device parameters in the product production process are time series data, that is, the data corresponding to the device parameters in the product production process are all arranged in time series.

[0034] Step 102: Based on the set of device parameters, determine the detection label of the product through a pre-trained product detection model, and the detection label is used to indicate whether the product is good or bad.

[0035] Among them, the product detection model can be any machine learning model, and the present invention does not limit the specific type of the product detection model.

[0036] It should be noted that the product detection model can be pre-trained. The product detection model can use the set of device parameters as the model input and the detection label used to indicate whether the product is good or bad as the model output to realize the prediction of whether the product is good or bad.

[0037] Step 103: Output the product detection result of the product based on the detection label of the product.

[0038] After determining the detection label used to indicate whether the product is good or bad, the determination of the product detection result can be realized based on the determined detection label.

[0039] The present invention obtains a set of device parameters composed of data corresponding to multiple device parameters generated in the product production process within a preset time period, and the data corresponding to each device parameter is arranged in time series. Thus, based on the set of device parameters, through a pre-trained product detection model, a detection label used to indicate whether the product is good or bad is determined, and then based on the detection label of the product, the product detection result of the product is output. The present invention can realize the prediction of whether the product is good or bad through a machine learning method, timely discover products that may be defective, and thus can timely give a warning to products that may be defective.

[0040] After introducing the basic implementation process of the present invention, the following introduces various optional implementation manners of the present invention.

[0041] In some embodiments, before step 101, step 100 of training a product detection model may further be included.

[0042] Optionally, the training process of the product detection model may include the following steps:

[0043] Step 100-1: Obtain training data, where the training data includes a set of sample device parameters corresponding to a sample product and a sample label of the sample product, and the sample label is used to indicate whether the sample product is good or bad.

[0044] In a possible implementation manner, a set of sample device parameters and detection data of multiple sample products may be obtained; based on a first threshold and the detection data of the multiple sample products, the sample label of each sample product is determined.

[0045] Optionally, these multiple sample products may be selected by a user according to actual technical requirements.

[0046] For example, a second screening interface may be provided, and the second screening interface is used to provide a function for setting product screening conditions, so that product screening conditions can be set through the second screening interface; based on the product screening conditions obtained through the second screening interface, multiple sample products are determined from the already produced products.

[0047] Optionally, the second screening interface may provide multiple optional screening conditions, and the user may set the optional screening conditions that meet the requirements according to actual technical requirements.

[0048] It should be noted that the optional screening conditions may include one or more of optional conditions such as product batch, production process, production equipment, production time, etc., but are not limited thereto. The screening conditions may also include more other types of conditions, and the specific types of the screening conditions are not limited in the present invention.

[0049] For example, referring to Figure 2 , Figure 2 is a schematic diagram of an interface of a second screening interface shown according to an embodiment of the present invention. As Figure 2 shown, the second screening interface provides a setting control corresponding to production time (i.e., Event Time), a setting control corresponding to manufacturer (i.e., Factory), a setting control corresponding to production process (i.e., Operation), a setting control corresponding to production equipment (i.e., Model), and a setting control corresponding to product type (i.e., Production Type). The user may set the screening parameters corresponding to each screening condition according to actual technical requirements, so as to implement the screening of sample products according to the screening parameters set by the user, and screen out multiple sample products from the already produced products.

[0050] It should be noted that after determining multiple sample products, the detection data of these determined sample products can be obtained.

[0051] Optionally, the detection data can be used to indicate the defect rate and / or the number of defects of the sample products, but is not limited thereto. The detection data can also include other types of data, and the present invention does not limit the data type of the detection data.

[0052] Optionally, the user can select the type of detection data required according to the actual technical needs. For example, the computing device can provide a detection data selection interface, and the detection data selection interface can provide various optional detection data so that the user can select from the various optional detection data provided by the detection data selection interface according to the actual technical needs.

[0053] See Figure 3 , Figure 3 is a schematic diagram of an interface of a detection data selection interface shown according to an embodiment of the present invention. As Figure 3 shown, the user can select specific result variables (i.e., detection data) according to the actual technical needs. See Figure 3 ,and the defect ratio can be used as the selected result variable.

[0054] It should be noted that there may be multiple different types of defects (or faults) on each sample product. Each sample product can correspond to multiple defect rates, and each defect rate can correspond to different types of defects.

[0055] Optionally, after the computing device obtains the corresponding type of detection data of multiple sample products according to the type of detection data selected by the user, it can determine the sample labels of each sample product based on the obtained detection data and the first threshold.

[0056] Among them, the first threshold can be determined based on the detection data of multiple sample products, or the first threshold can be obtained through the first screening interface provided by the computing device.

[0057] In a possible implementation manner, when determining the first threshold based on the detection data of multiple sample products, the focusing threshold can be calculated by an algorithm, and the calculated focusing threshold is the first threshold.

[0058] Optionally, when calculating the focusing threshold by an algorithm to obtain the first threshold, it can be implemented through the following steps:

[0059] Step 1: Sort the detection data of multiple sample products to obtain a to-be-processed array.

[0060] Optionally, the detection data of all sample products can be sorted in ascending order of values to obtain an array to be processed, which can be denoted as SortedData.

[0061] Step 2: Based on the median of the array to be processed, divide the array to be processed into a first array and a second array.

[0062] Among them, the median can be the value in the middle position after a set of data is arranged in ascending order. For example, if the data included in the array to be processed are arranged in ascending order as 3, 5, 7, 9, 11, it can be seen that the value in the middle position is 7, so 7 is the median of the array to be processed.

[0063] It should be noted that when the number of data in the array to be processed is odd, the value in the middle position after the data in the array to be processed are arranged in ascending order is one, and this value is the median of the array to be processed; when the number of data in the data to be processed is even, the values in the middle position after the data in the array to be processed are arranged in ascending order are two, and the average value of these two values is the median of the array to be processed.

[0064] The median can divide a set of data into two parts. Among them, some data are all larger than the median, while the other part of the data are all smaller than the median.

[0065] Therefore, the data in the array to be processed whose positions are before the median of the array to be processed can be divided into a first array, that is, the data whose values are less than the median of the array to be processed can be divided into a first array, and the first array can be denoted as LowerGroup; the data in the array to be processed whose positions are after the median of the array to be processed can be divided into a second array, that is, the data whose values are greater than the median of the array to be processed can be divided into a second array, and the second array can be denoted as UpperGroup.

[0066] Step 3: Determine the first average value of the detection data included in the first array and the second average value of the detection data included in the second array respectively.

[0067] That is, the average values of the detection data included in the first array LowerGroup and the second array UpperGroup can be calculated respectively. For example, the sum value of the detection data included in the first array LowerGroup can be calculated, and then the ratio of the calculated sum value to the number of detection data included in the first array LowerGroup can be determined as the first average value; the sum value of the detection data included in the second array UpperGroup can be calculated, and then the ratio of the calculated sum value to the number of detection data included in UpperGroup can be determined as the second average value.

[0068] Optionally, the first mean value of the detection data included in the first array LowerGroup can be denoted as Mean l , and the second mean value of the detection data included in the second array UpperGroup can be denoted as Mean u .

[0069] Step Four: Determine the absolute value of the difference between the detection data of each sample product and the first mean value to obtain a first difference array, and determine the absolute value of the difference between the detection data of each sample product and the second mean value to obtain a second difference array.

[0070] Optionally, each detection data included in the array to be processed SortedData can be subtracted from the first mean value Mean l and the absolute value is taken to obtain a first difference array, which can be denoted as DiffLowerMean; each detection data included in the array to be processed SortedData can be subtracted from the second mean value Mean u and the absolute value is taken to obtain a second difference array, which can be denoted as DiffUpperMean.

[0071] Step Five: Determine the target position parameter based on the data at the corresponding positions in the first difference array and the second difference array.

[0072] In a possible implementation, the data at the corresponding positions in the first difference array and the second difference array can be compared one by one, and the number of positions where the data at a certain position in the first difference array is greater than the data at the corresponding position in the second difference array can be determined as the target position parameter.

[0073] Optionally, the data at the corresponding positions in the first difference array DiffLowerMean and the second difference array DiffUpperMean can be compared one by one. If DiffLowerMean[i] < DiffUpperMean[i], then record that index is incremented by one. DiffLowerMean[i] represents the data at the i-th position in the first difference array, and DiffUpperMean[i] represents the data at the i-th position in the second difference array. The initial value of index can be 0, and the updated value of index can be denoted as CurrentIndex; repeat the above steps iteratively until CurrentIndex remains unchanged, and then the CurrentIndex when it remains unchanged can be used as the target position parameter.

[0074] Step Six: Determine the first threshold based on the target position parameter and the array to be processed.

[0075] In a possible implementation, it is possible to determine the data at the first target position indicated by the target position parameter in the array to be processed, and the data at the second target position corresponding to the first target position, where the second target position is the position before the first target position; based on the average value of the data at the first target position and the data at the second target position, determine the first threshold value.

[0076] Optionally, the average value of the data at the first target position and the data at the second target position can be determined as the first threshold value.

[0077] For example, the first threshold value can be calculated according to the following formula (1):

[0078] Focus = (SortedData[CurrentIndex] + SortedData[CurrentIndex - 1]) / 2 (1)

[0079] Where Focus represents the first threshold value, SortedData[CurrentIndex] represents the data at the position indicated by the target position parameter in the array to be processed, and SortedData[CurrentIndex - 1] represents the data at the position before the position indicated by the target position parameter in the array to be processed.

[0080] Optionally, when obtaining the first threshold value through the first screening interface provided by the computing device, a threshold setting control can be provided in the first screening interface so that the user can set the first threshold value through the threshold setting control.

[0081] It should be noted that a sample distribution diagram can be drawn based on the detection data of multiple sample products, and the user can set the first threshold value according to the actual situation of the sample distribution diagram and in combination with relevant experience judgments.

[0082] In more possible implementations, the computing device can draw a chart or a data visualization diagram (such as a Chart diagram) based on the detection data of multiple sample products. Among them, the horizontal axis of the Chart diagram can select Time, Context, and Tracking Time to implement the drawing of the sample distribution diagram. See Figure 4 , Figure 4 is a schematic diagram of a sample distribution diagram shown according to an embodiment of the present invention.

[0083] It should be noted that Time refers to events or processes that occur at a specific moment or within a specific time period. In data analysis and data visualization, time is usually an important dimension for analyzing and understanding data. Through the analysis of the time dimension, trends, periodicity, seasonality, and other characteristics of data changes over time can be revealed. Context refers to the relevant background information or conditions considered when analyzing or interpreting data. The context can include factors such as the data collection environment, data sources, data uses, data limitations, etc. Considering the context can help better understand the meaning and influencing factors of data and avoid misinterpreting or misleading the data. Tracking Time refers to recording and tracking the time when specific events or processes occur. In data analysis and monitoring, tracking time can help us understand the order of occurrence, duration, and relationships between events. Through effective time tracking, we can conduct event backtracking, analysis, and prediction, and thus make corresponding decisions and optimizations. Optionally, a certain process under Tracking can be selected to redraw the Chart according to the time corresponding to the selected process, facilitating the user to clearly analyze the positive and negative sample distributions during this process for filtering and classification).

[0084] After obtaining the first threshold through the above embodiments, the sample labels of each sample product can be determined based on the first threshold and the detection data of multiple sample products.

[0085] However, it should be noted that not all sample products can be used as training data. In more possible implementation manners, multiple sample products determined based on product screening conditions can also be filtered based on preset filtering parameters to obtain training data based on the filtered sample products.

[0086] Among them, the filtering parameters can be used for filtering based on detection data. Optionally, the user can input a filtering range value according to actual technical requirements, so that the computing device can implement sample filtering based on the filtering range value input by the user to filter and delete some samples with no reference value, improving the reliability of sample analysis.

[0087] Alternatively, the filtering parameters can be used for filtering based on the arrival rate. Optionally, the user can input an arrival rate range (that is, the sampling ratio range that meets the technical requirements) according to actual technical requirements, so that the computing device can implement sample filtering based on the arrival rate range to filter and delete some samples with low reference value due to a small number of samples selected for inspection, improving the reliability of sample analysis.

[0088] Alternatively, filtering parameters can be used to filter based on production equipment. Optionally, the user can select the equipment identifier of the production equipment corresponding to the samples to be retained according to actual technical requirements, so that the computing device can filter and delete some samples that are not produced by the production equipment selected by the user based on the equipment identifier of the production equipment selected by the user, improving the pertinence of sample analysis.

[0089] It should be noted that the process of filtering the multiple sample products determined based on the product screening conditions according to the preset filtering parameters described above can be an optional step. In more embodiments, this step can be omitted or replaced.

[0090] Optionally, when determining the sample labels of each sample product based on the first threshold and the detection data of multiple sample products, it can be implemented in the following manner:

[0091] In a possible implementation manner, the sample label of the sample product whose detection data is less than the first threshold is determined as the first sample label, and the first sample label is used to indicate that the sample product is good.

[0092] Optionally, the sample products whose detection data is less than the first threshold can be used as normal defect-free samples (i.e., positive samples). The sample labels of such samples are the first sample labels, and the first sample labels can be used to indicate that the sample products are good.

[0093] In another possible implementation manner, the sample label of the sample product whose detection data is greater than or equal to the first threshold is determined as the second sample label, and the second sample label is used to indicate that the sample product is defective.

[0094] Optionally, the sample products whose detection data is greater than or equal to the first threshold can be used as defective samples (i.e., negative samples). The sample labels of such samples are the second sample labels, and the second sample labels can be used to indicate that the sample products are defective.

[0095] Optionally, when obtaining the first threshold through the first screening interface, in addition to setting the first threshold through the first screening interface, the user can also set the division method of positive and negative samples through the first screening interface. For example, the first screening interface can provide multiple optional division methods such as >, <, <>, ><, etc., so that the user can set the screening interval to which the detection data of the sample products to be used as positive samples belongs according to actual technical requirements. The sample products corresponding to the detection data within the screening interval are the sample products to be used as training samples.

[0096] After the positive and negative samples are divided through the above embodiments, the user can, according to actual technical requirements, set the process stations and equipment (i.e., Operation) that may cause defects through the third screening interface, so as to obtain the set of equipment parameters reported by the process stations and equipment that may cause defects.

[0097] See Figure 5 , Figure 5 is a schematic diagram of an interface of a third screening interface shown according to an embodiment of the present invention. As Figure 5 shown, the batch identification (i.e., Lot ID) and product identification (i.e., Glass ID) of multiple sample products can be used as sample identifiers, and positive and negative sample labels can be marked for each sample product. Among them, Y can represent a positive sample, and N can represent a negative sample. In the third screening interface as Figure 5 shown, the user can select the site / equipment identifier to be analyzed (i.e., Operation ID) and the identifier of the set of equipment parameters of specific sub-equipment or processing units in the site / equipment (i.e., Recipe ID), so as to obtain the set of equipment parameters (i.e., Recipe) reported by the process stations and equipment that may cause defects. The set of equipment parameters can include various types of SV parameters, including but not limited to temperature, flow rate, and pressure, etc.

[0098] Step 100-2: Extract features based on the set of sample equipment parameters to obtain sample input features.

[0099] Optionally, the acquisition of sample input features can be realized by a method of Time Series Feature Extraction on basis of Scalable Hypothesis tests (Tsfresh).

[0100] It should be noted that the Tsfresh method is a method for feature extraction of time series data, which can automatically extract relevant features of time series data. Since the data corresponding to the equipment parameters included in the set of sample equipment parameters is time series data, the Tsfresh method can be used to extract its features.

[0101] In a possible implementation manner, step 100-2 can be realized through the following steps:

[0102] Step 1: Obtain various sample statistical features of multiple sample equipment parameters included in the set of sample equipment parameters.

[0103] Among them, the sample statistical features may include sample mean, sample maximum value, sample correlation, physics-based non-linearity and complexity indicators, data signal processing-related indicators, etc. The present invention does not limit the feature types of the sample statistical features.

[0104] It should be noted that the Tsfresh method can provide an application programming interface (API) for automatically generating features. Through this interface, the Tsfresh method can be called to generate hundreds of sample statistical features from one time series variable, including but not limited to descriptive statistical features (mean, maximum value, correlation, etc.), physics-based non-linearity and complexity indicators, digital signal processing-related indicators, etc.

[0105] Step 2: Perform feature significance tests on multiple sample statistical features to determine the importance parameters corresponding to each sample statistical feature. The importance parameters are used to indicate the influence degree of the corresponding sample statistical feature on the sample label.

[0106] Optionally, based on hypothesis testing, the importance of each sample statistical feature to the prediction target under study can be evaluated separately and independently. For example, significance tests can be performed on each sample statistical feature to determine their contributions to the prediction target. The significance test methods include but are not limited to the t-test method for testing whether there are significant differences in the means of two groups of samples, the analysis of variance (ANOVA) method for testing whether there are significant differences in the means of multiple groups of samples, and the Pearson correlation coefficient for measuring the strength of the linear relationship between two variables.

[0107] It should be noted that the importance can be quantitatively represented as a p-value to quantify the importance of each sample statistical feature to the prediction label.

[0108] Step 3: Perform feature selection among multiple sample statistical features according to a preset second threshold and the importance parameters corresponding to the multiple sample statistical features to obtain sample input features.

[0109] Among them, the second threshold can be determined based on the values of the importance parameters corresponding to the multiple sample statistical features. The second threshold can take any value, and the present invention does not limit the specific value of the second threshold.

[0110] It should be noted that according to the preset second threshold, sample statistical features with corresponding importance parameters less than the second threshold can be filtered out to delete features with low correlation to the prediction target, and only relatively important features are used as input to prevent the occurrence of the curse of dimensionality.

[0111] Optionally, when performing feature selection, data with empty statistical feature values can also be removed.

[0112] Optionally, before performing feature extraction, data preprocessing can also be performed on the multiple sample device parameters included in the sample device parameter set. The present invention does not limit the specific data preprocessing method.

[0113] Step 100-3: Perform model training based on the sample input features and sample labels to obtain a product detection model.

[0114] Optionally, the sample input features can be used as the model input, and the sample prediction labels can be output through model processing. Then, based on the difference between the sample prediction labels and the samples, model training is performed to obtain a product detection model.

[0115] Among them, the product detection model can be a binary classification model. For example, the product detection model can be a logistic regression model, and / or the product detection model can be a decision tree, but it is not limited thereto. The product detection model can also be other types of models.

[0116] Among them, the logistic regression model can use the Sigmoid function as the activation function to map the continuous results of linear regression to discrete values, and use the mapped discrete values as the prediction probabilities (that is, the probabilities that the samples belong to a certain category). Then, based on the prediction probabilities, it is determined whether the samples belong to a certain category. Compared with linear regression, the ability to solve non-linear problems is increased.

[0117] In addition, it should be noted that the logistic regression model can use the gradient descent optimization algorithm. By continuously reducing the value of the loss function, the weight coefficients of the corresponding mapping algorithm in the logistic regression model are continuously updated to increase the probability that the sample belongs to the positive category and reduce the probability that the sample belongs to the negative category.

[0118] The decision tree is composed of a root node, several internal nodes, and leaf nodes. Among them, the leaf nodes correspond to the classification results, and the other nodes correspond to the intermediate processing results of the attribute judgment rules. It is necessary to make judgments recursively layer by layer according to the conditions from the root node to the leaf nodes to finally realize the prediction of the sample labels.

[0119] The training of the decision tree is essentially to find in what order of features to perform recursive judgments to find the purest division nodes, so as to find the features that can best distinguish different samples as the priority judgment conditions, thereby improving the purity of the decision tree.

[0120] Optionally, a Classification and Regression Tree (CART) can be used as the decision tree, and the Gini coefficient can be used as the splitting rule. The Gini coefficient can be used to represent the influence degree and importance of each feature on the prediction result. The smaller the Gini coefficient, the lower the uncertainty of the prediction result, and it is more optimal as a splitting point. In a K-classification problem, for a given sample set D, its Gini coefficient is where C k is the subset of samples in the sample set D that belong to the k-th class.

[0121] When dealing with classification problems through decision trees, there is no need to normalize the data, which reduces the data preprocessing work. Moreover, variables can be automatically screened, the decision-making process is easy to interpret, and the visualization difficulty of the decision-making process is relatively small.

[0122] It should be noted that the above are only two exemplary implementation manners of the product detection model, but are not limited thereto.

[0123] In addition, it should be noted that when training the product detection model, a supervised training method can be adopted. For example, 80% of the finally obtained training data can be used to train the model, and the remaining 20% can be used as test data for verifying the prediction accuracy of the model.

[0124] Optionally, data related to the feature importance ranking (such as the Gini coefficient) can be generated during the model training process, and the data related to the feature importance ranking can be output and visually displayed, so that users can evaluate the importance of different device parameters based on the data related to the feature importance ranking.

[0125] Taking the product detection model for training the detection label for identifying OLED products as an example, the model training process provided in the above embodiments can be referred to Figure 6 , Figure 6 which is a flowchart of a model training process shown according to an embodiment of the present invention, as Figure 6As shown, first, determine multiple sample products among the already produced OLED products based on product screening conditions to achieve sample creation; then, divide the positive and negative samples based on the determined multiple sample products to achieve sample screening; then, obtain the set of device parameters based on the screened sample products, and thus, through an algorithm model, perform data preprocessing on the obtained set of device parameters, and then perform a series of processes such as Tsfresh feature extraction and normalization processing based on the preprocessed set of device parameters to achieve feature selection, so as to obtain the sample input features for input into the model, and thus, the training of the decision tree or logistic regression model can be achieved. During the model training process, data related to the sorting of feature importance can be generated. After the model training is completed, the trained model can be saved. When saving, the user can choose whether to save the decision tree model or the logistic regression model, and thus save the model selected by the user to the database. And, a function to view the historical records of the models saved by the user can also be provided.

[0126] After completing the training of the product detection model through the above process, the prediction of whether the product is good or bad can be achieved based on the trained product detection model.

[0127] In some embodiments, for step 102, when determining the detection label of the product based on the set of device parameters through a pre-trained product detection model, it can be achieved through the following steps:

[0128] Step 1021: Extract features based on the set of device parameters to obtain input features.

[0129] Optionally, obtain various statistical features of the multiple device parameters included in the set of device parameters; perform feature selection among the various statistical features according to a preset second threshold to obtain input features.

[0130] For the specific implementation process of step 1021, reference can be made to the above embodiments, and details will not be elaborated here.

[0131] Step 1022: Input the input features into the product detection model and output the detection label of the product.

[0132] In some embodiments, after obtaining the detection label of the product, through step 103, the product detection result of the product can be output based on the detection label of the product.

[0133] Optionally, the detection label of the product can be directly output as the product detection result. That is, if the detection label of the product indicates that the product is good, the product detection result of the product can be output as good; if the detection label of the product indicates that the product is bad, the product detection result of the product can be output as bad.

[0134] In addition, it should be noted that for products with poor product detection results, alarm information (including but not limited to audio alarm) can be sent in a timely manner so that users can be informed of the occurrence of product defects in a timely manner.

[0135] It should be noted that the process of obtaining detection labels through the above model can be real-time, so as to effectively monitor the production process of products, timely warn products that may have defects, and effectively avoid the occurrence of defects.

[0136] Through the product defect detection method provided by the present invention, on the one hand, users can view the sorting of feature importance to help users locate the position where product defects occur; on the other hand, the model obtained through training can be used to predict the defects of newly produced products, so as to timely discover product defect problems, and thus timely warning can be carried out to avoid the occurrence of more defects. Therefore, the product defect detection method provided by the present invention can improve the analysis efficiency of business personnel, save labor costs, improve detection and repair efficiency, and assist in improving product yield.

[0137] Corresponding to the embodiments of the foregoing method, the present invention also provides embodiments of a product defect detection device and a computing device to which it is applied.

[0138] As Figure 7 shown, Figure 7 is a block diagram of a product defect detection device shown according to an embodiment of the present invention. The device includes:

[0139] An acquisition module 701, configured to acquire a set of device parameters during the product production process within a preset time period. The set of device parameters includes data corresponding to multiple device parameters. For any device parameter, the data corresponding to the device parameter is arranged in time sequence;

[0140] A determination module 702, configured to determine a detection label of the product based on the set of device parameters through a pre-trained product detection model. The detection label is used to indicate whether the product is good or defective;

[0141] An output module 703, configured to output a product detection result of the product based on the detection label of the product.

[0142] In some embodiments, when the determination module 702 is configured to determine a detection label of the product based on the set of device parameters through a pre-trained product detection model, it is configured to:

[0143] Extract features based on the set of device parameters to obtain input features;

[0144] Input the input features into the product detection model and output the detection label of the product.

[0145] In some embodiments, the determining module 702, when used for feature extraction based on a set of device parameters to obtain input features, is configured to:

[0146] Obtain various statistical features of multiple device parameters included in the set of device parameters;

[0147] According to a preset second threshold, perform feature selection among the various statistical features to obtain input features.

[0148] In some embodiments, the apparatus further includes a training module, configured to:

[0149] Obtain training data, where the training data includes a set of sample device parameters corresponding to a sample product and a sample label of the sample product, and the sample label is used to indicate whether the sample product is good or bad;

[0150] Perform feature extraction based on the set of sample device parameters to obtain sample input features;

[0151] Perform model training based on the sample input features and the sample labels to obtain a product detection model.

[0152] In some embodiments, when the training module is used for obtaining training data, it is configured to:

[0153] Obtain a set of sample device parameters and detection data of multiple sample products, where the detection data is used to indicate the defective rate and / or the number of defective products of the sample products;

[0154] Based on the first threshold and the detection data of multiple sample products, determine the sample labels of each sample product.

[0155] In some embodiments, the training module is further configured to determine the first threshold;

[0156] When the training module is used for determining the first threshold, it is configured to perform any one of the following:

[0157] Based on the detection data of multiple sample products, determine the first threshold;

[0158] Provide a first screening interface, and obtain the first threshold through the first screening interface.

[0159] In some embodiments, when the training module is used for determining the first threshold based on the detection data of multiple sample products, it is configured to:

[0160] Sort the detection data of multiple sample products to obtain an array to be processed;

[0161] Based on the median of the array to be processed, divide the array to be processed into a first array and a second array;

[0162] Determine the first mean value of the detection data included in the first array and the second mean value of the detection data included in the second array respectively;

[0163] Determine the absolute value of the difference between the detection data of each sample product and the first mean value to obtain a first difference array, and determine the absolute value of the difference between the detection data of each sample product and the second mean value to obtain a second difference array;

[0164] Based on the data at the corresponding positions in the first difference array and the second difference array, determine the target position parameter;

[0165] Based on the target position parameter and the array to be processed, determine the first threshold.

[0166] In some embodiments, when the training module is used to determine the first threshold based on the target position parameter and the array to be processed, it is used to:

[0167] Determine the data at the first target position indicated by the target position parameter in the array to be processed and the data at the second target position corresponding to the first target position, where the second target position is the previous position of the first target position;

[0168] Based on the mean value of the data at the first target position and the data at the second target position, determine the first threshold.

[0169] In some embodiments, when the training module is used to determine the sample labels of each sample product based on the first threshold and the detection data of multiple sample products, it is used to:

[0170] Determine the sample label of the sample product with detection data less than the first threshold as the first sample label, and the first sample label is used to indicate that the sample product is good;

[0171] Determine the sample label of the sample product with detection data greater than or equal to the first threshold as the second sample label, and the second sample label is used to indicate that the sample product is defective.

[0172] In some embodiments, the training module is further used to:

[0173] Provide a second screening interface, and the second screening interface is used to provide the function of setting product screening conditions;

[0174] Based on the product screening conditions obtained through the second screening interface, determine multiple sample products from the products that have been produced.

[0175] In some embodiments, the training module is further used for any one of the following:

[0176] Based on the preset filtering parameters, filter the multiple sample products determined based on the product screening conditions, so as to obtain training data based on the filtered sample products;

[0177] Among them, the filtering parameter is used to filter based on the detection data, or the filtering parameter is used to filter based on the arrival rate, or the filtering parameter is used to filter based on the production equipment.

[0178] In some embodiments, when the training module is used to extract features based on the sample device parameter set to obtain sample input features, it is used for:

[0179] Obtain various sample statistical features of multiple sample device parameters included in the sample device parameter set;

[0180] Perform a feature significance test on various sample statistical features to determine the importance parameter corresponding to each sample statistical feature, where the importance parameter is used to indicate the influence degree of the corresponding sample statistical feature on the sample label;

[0181] According to a preset second threshold and the importance parameters corresponding to various sample statistical features, perform feature selection among various sample statistical features to obtain sample input features.

[0182] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial descriptions of the method embodiments. The device embodiments described above are only illustrative. The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules, that is, they may be located in one place, or may be distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution in this specification. Those of ordinary skill in the art can understand and implement it without creative work.

[0183] The present invention also provides a computing device. Refer to Figure 8 , Figure 8 is a schematic structural diagram of a computing device shown according to an embodiment of the present invention. As Figure 8 shown, the computing device includes a processor 810, a memory 820, and a network interface 830. The memory 820 is used to store computer instructions that can run on the processor 810. The processor 810 is used to implement the product defect detection method provided by any embodiment of the present invention when executing the computer instructions. The network interface 830 is used to implement input and output functions. In more possible implementation manners, the computing device may further include other hardware, and the present invention does not limit this.

[0184] The present invention also provides a computer-readable storage medium. The computer-readable storage medium can be in various forms. For example, in different examples, the computer-readable storage medium can be: RAM (Random Access Memory), volatile memory, non-volatile memory, flash memory, storage drives (such as hard disk drives), solid-state drives, any type of storage disk (such as optical discs, DVDs, etc.), or similar storage media, or a combination thereof. Specifically, the computer-readable medium can also be paper or other suitable media capable of printing programs. A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, it implements the product defect detection method provided by any embodiment of the present invention.

[0185] The present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the product defect detection method provided by any embodiment of the present invention.

[0186] Those skilled in the art should understand that one or more embodiments of this specification can be provided as a method, an apparatus, a computing device, a computer-readable storage medium, or a computer program product. Therefore, one or more embodiments of this specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0187] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the embodiment corresponding to the computing device, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment.

[0188] The above describes specific embodiments of this specification. Other embodiments are within the scope of the present invention. In some cases, the actions or steps recorded in the present invention can be executed in a different order from that in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0189] The embodiments of the subject matter and the functional operations described in this specification can be implemented in the following: digital electronic circuits, tangible computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or a combination of one or more of them. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules in computer program instructions encoded on a tangible non-transitory program carrier to be executed by or to control the operation of a product defect detection device. Alternatively or additionally, the program instructions can be encoded on a manually generated propagated signal, such as a machine-generated electrical, optical, or electromagnetic signal, which is generated to encode and transmit information to a suitable receiver device for execution by the product defect detection device. A computer storage medium can be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.

[0190] The processes and logical flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform the corresponding functions by operating on input data and generating output. The processes and logical flows can also be performed by special-purpose logic circuitry, such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit), and the apparatus can also be implemented as special-purpose logic circuitry.

[0191] Computers suitable for executing computer programs include, for example, general and / or special-purpose microprocessors, or any other type of central processing unit. Generally, the central processing unit will receive instructions and data from a read-only memory and / or a random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Generally, a computer will also include one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, etc., or the computer will be operably coupled to such mass storage devices to receive data from them or transmit data to them, or both. However, a computer is not necessarily required to have such devices. In addition, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name just a few.

[0192] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, including, for example, semiconductor memory devices (such as EPROM, EEPROM, and flash memory devices), magnetic disks (such as internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. The processor and the memory may be supplemented by, or incorporated in, special purpose logic circuitry.

[0193] Although this specification contains many specific implementation details, these should not be construed as limiting the scope of any invention or the scope of what is claimed, but rather as mainly describing the features of specific embodiments of a particular invention. Certain features that are described in multiple embodiments in this specification may also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment may also be implemented separately in multiple embodiments or in any suitable sub-combination. In addition, although features may operate in certain combinations as described above and even be initially claimed as such, one or more features from a claimed combination may in some cases be removed from that combination, and the claimed combination may be directed to a sub-combination or a variation of a sub-combination.

[0194] Similarly, although operations are depicted in the drawings in a particular order, this should not be construed as requiring that the operations be performed in the particular order shown or sequentially, or that all illustrated operations be performed, to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of the various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0195] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the present invention. In some cases, the acts recited in the present invention may be performed in a different order and still achieve the desired result. In addition, the processes depicted in the drawings are not necessarily in the particular order or sequential order shown to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0196] Those skilled in the art will readily conceive of other embodiments of this specification after considering the specification and practicing the invention as claimed herein. This specification is intended to cover any variations, uses, or adaptations of this specification that follow the general principles of this specification and include common general knowledge or conventional technical means in the technical field not claimed in this application. That is, this specification is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope.

[0197] The above are only optional embodiments of this specification and are not intended to limit this specification. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this specification shall be included within the scope of protection of this specification.

[0198] It should be noted that the forming processes adopted by the processes involved in the present invention may include, for example, film-forming processes such as deposition and sputtering, and patterning processes such as etching.

[0199] In the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance. The term "plurality" means two or more, unless otherwise clearly defined.

[0200] Those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the disclosure herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include known common general knowledge or conventional technical means in the technical field not disclosed by the present invention. The specification and examples are only to be regarded as exemplary, and the true scope and spirit of the present invention are pointed out by the following claims.

[0201] It should be understood that the present invention is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A method for detecting product defects, characterized in that, The method includes: Obtaining a set of device parameters during the product production process within a preset time period, where the set of device parameters includes data corresponding to multiple device parameters. For any device parameter, the data corresponding to the device parameter is arranged in time sequence; Based on the set of device parameters, determining a detection label of the product through a pre-trained product detection model, where the detection label is used to indicate whether the product is good or bad; Based on the detection label of the product, outputting a product detection result of the product.

2. The method according to claim 1, characterized in that, The determining the detection label of the product through the pre-trained product detection model based on the set of device parameters includes: Performing feature extraction based on the set of device parameters to obtain input features; Inputting the input features into the product detection model and outputting the detection label of the product.

3. The method according to claim 2, wherein The performing feature extraction based on the set of device parameters to obtain input features includes: Obtaining various statistical features of multiple device parameters included in the set of device parameters; Performing feature selection among the various statistical features according to a preset second threshold to obtain the input features.

4. The method according to claim 1, characterized in that, The training process of the product detection model includes: Obtaining training data, where the training data includes a set of sample device parameters corresponding to a sample product and a sample label of the sample product, and the sample label is used to indicate whether the sample product is good or bad; Performing feature extraction based on the set of sample device parameters to obtain sample input features; Performing model training based on the sample input features and the sample label to obtain the product detection model.

5. The method according to claim 4, wherein The obtaining the training data includes: Obtaining a set of sample device parameters and detection data of multiple sample products, where the detection data is used to indicate the defect rate and / or the number of defects of the sample products; Based on a first threshold and the detection data of the multiple sample products, determining the sample label of each sample product.

6. The method according to claim 5, characterized in that The determining process of the first threshold includes any one of the following: Based on the detection data of the multiple sample products, determining the first threshold; Providing a first screening interface and obtaining the first threshold through the first screening interface.

7. The method according to claim 6, wherein The determining the first threshold based on the detection data of the multiple sample products includes: Sorting the detection data of the multiple sample products to obtain an array to be processed; Based on the median of the array to be processed, dividing the array to be processed into a first array and a second array; Respectively determining a first mean value of the detection data included in the first array and a second mean value of the detection data included in the second array; Determining the absolute value of the difference between the detection data of each sample product and the first mean value to obtain a first difference array, and determining the absolute value of the difference between the detection data of each sample product and the second mean value to obtain a second difference array; Based on the data at the corresponding positions in the first difference array and the second difference array, determining a target position parameter; Based on the target position parameter and the array to be processed, determining the first threshold.

8. The method according to claim 7, wherein The determining the first threshold based on the target position parameter and the array to be processed includes: Determine the data at the first target position indicated by the target position parameter in the array to be processed, and the data at the second target position corresponding to the first target position, where the second target position is the previous position of the first target position; Determine the first threshold based on the average value of the data at the first target position and the data at the second target position.

9. The method according to claim 5, characterized in that, The determining the sample labels of each sample product based on the first threshold and the detection data of the multiple sample products includes: Determine the sample label of the sample product with detection data less than the first threshold as the first sample label, where the first sample label is used to indicate that the sample product is good; Determine the sample label of the sample product with detection data greater than or equal to the first threshold as the second sample label, where the second sample label is used to indicate that the sample product is defective.

10. The method according to claim 4, wherein Before obtaining the sample device parameter sets and detection data of the multiple sample products, the method further includes: Provide a second screening interface, where the second screening interface is used to provide a function for setting product screening conditions; Determine the multiple sample products from the products that have been produced based on the product screening conditions obtained through the second screening interface.

11. The method according to claim 10, wherein The method further includes any one of the following: Filter the multiple sample products determined based on the product screening conditions based on a preset filtering parameter, so as to obtain the training data based on the filtered sample products; Wherein, the filtering parameter is used to filter based on the detection data, or the filtering parameter is used to filter based on the arrival rate, or the filtering parameter is used to filter based on the production equipment.

12. The method according to claim 4, wherein The extracting features based on the sample device parameter sets to obtain sample input features includes: Obtain various sample statistical features of the multiple sample device parameters included in the sample device parameter sets; Perform a feature significance test on the various sample statistical features to determine the importance parameter corresponding to each sample statistical feature, where the importance parameter is used to indicate the influence degree of the corresponding sample statistical feature on the sample label; Perform feature selection among the various sample statistical features according to a preset second threshold and the importance parameters corresponding to the various sample statistical features to obtain the sample input features.

13. A defective product detection device, characterized in that, The device includes: An acquisition module, configured to acquire a set of device parameters during the product production process within a preset time period, where the set of device parameters includes data corresponding to multiple device parameters, and for any device parameter, the data corresponding to the device parameter is arranged in time sequence; A determination module, configured to determine the detection label of the product based on the set of device parameters through a pre-trained product detection model, where the detection label is used to indicate that the product is good or defective; An output module, configured to output the product detection result of the product based on the detection label of the product.

14. A computing device, characterized in that, The computing device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. Wherein, when the processor executes the computer program, the operations performed by the product defect detection method according to any one of claims 1 to 12 are implemented.

15. A computer-readable storage medium, characterized in that, A program is stored on the computer-readable storage medium. When the program is executed by a processor, the operations performed by the product defect detection method according to any one of claims 1 to 12 are implemented.

Images