CN115640933A - Method, device and equipment for automatically managing production line defects and storage medium - Google Patents
Method, device and equipment for automatically managing production line defects and storage medium Download PDFInfo
- Publication number
- CN115640933A CN115640933A CN202211368271.3A CN202211368271A CN115640933A CN 115640933 A CN115640933 A CN 115640933A CN 202211368271 A CN202211368271 A CN 202211368271A CN 115640933 A CN115640933 A CN 115640933A
- Authority
- CN
- China
- Prior art keywords
- state
- neural network
- equipment
- production line
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to the field of production line manufacturing, in particular to a method, a device, equipment and a storage medium for automatically managing production line defects. It includes: acquiring the equipment state corresponding to the production line equipment when the product manufactured by the production line has defects; the equipment state comprises control instructions and process parameters of production line equipment; acquiring a pre-trained state optimization model; inputting the equipment state into a state optimization model to obtain an optimized equipment state of the production line equipment; and inputting the optimized equipment state into the production line equipment to optimize the manufacturing defects of the production line equipment. The defect of processing equipment in a production line can be analyzed and detected in a manual mode, optimization is carried out, a large amount of time and energy are consumed, and the problem that the production line defect management efficiency is low is solved.
Description
Technical Field
The application relates to the field of production line manufacturing, in particular to a method, equipment and a storage medium for automatically managing production line defects.
Background
Complex production lines often include tens or even hundreds of production steps, for example, semiconductor manufacturing lines typically require hundreds of stages. The complex production line is often equipped with fine processing equipment, and if the control instruction of the processing equipment or the setting of the process parameters has defects, the products manufactured by the production line also have defects, so that the optimization of the defects of the processing equipment in the production line is a main direction of industry development.
A conventional method for optimizing defects in processing equipment in a production line includes: the method comprises the steps of analyzing and detecting defects of processing equipment in a production line in a manual mode, and then manually adjusting control instructions or technological parameters of the processing equipment with the defects so as to optimize the processing equipment.
However, analyzing and detecting defects of processing equipment in a production line in a manual manner and performing optimization require a lot of time and effort, and there is a problem in that the efficiency of managing defects in the production line is low.
Disclosure of Invention
The application provides a method, a device, equipment and a storage medium for automatically managing production line defects, which can solve the problem that the defects of processing equipment in a production line are analyzed and detected in a manual mode, optimize the defects, consume a large amount of time and energy, and have the problem of low production line defect management efficiency, and the application provides the following technical scheme:
in a first aspect, a method for automatically managing production line defects is provided, which includes: acquiring the equipment state corresponding to the production line equipment when the product manufactured by the production line has defects; the equipment state comprises control instructions and process parameters of the production line equipment; acquiring a pre-trained state optimization model; inputting the equipment state into the state optimization model to obtain an optimized equipment state of the production line equipment; inputting the optimized equipment state into the production line equipment to optimize the manufacturing defects of the production line equipment; the obtaining of the pre-trained state optimization model includes: inputting the equipment states of the Nth group of samples into a first neural network model to obtain an Nth group of sample adjustment states corresponding to the equipment states of the Nth group of samples; n is an integer greater than or equal to 1; the sample device states comprise K groups of sample device states, wherein K is an integer greater than or equal to 1; inputting the adjustment state of the Nth group of samples into a pre-trained defect prediction model to obtain a prediction result of the Nth group of samples; inputting the Nth group of sample equipment states, the Nth group of sample adjustment states and the Nth group of sample prediction results into a preset reward function to obtain an Nth reward value; adjusting the first neural network model based on the Nth reward value to obtain an adjusted Nth first neural network model; inputting N = N +1, inputting the state of the N +1 group of sample devices into the Nth first neural network model, and repeating the steps until N is equal to a preset accumulated step length; under the condition that the N is equal to the preset accumulated step length, determining a first neural network model meeting a preset reward condition from N first neural network models; and repeating the training steps on the first neural network model which meets the preset reward condition until a preset convergence condition or a preset training frequency is reached, and taking the obtained current first neural network model as a state optimization model.
Optionally, the inputting N = N +1 and the N +1 th group of sample device states into the nth first neural network model, and repeating the above steps until N is equal to a preset cumulative step size includes: when N is equal to K and K is smaller than a preset accumulated step length, acquiring a variable M, wherein the variable M is used for recording the times of N = 1; taking the Nth first neural network model as a first neural network model to be input; inputting the N =1, M = M +1, and the Nth group of sample device states into the first neural network model to be input, so as to obtain an (M-1) K + N group of sample adjustment states corresponding to the Nth group of sample device states; inputting the (M-1) K + N groups of sample adjustment states into a pre-trained defect prediction model to obtain an (M-1) K + N group of sample prediction results; inputting the device state of the Nth group of samples, the adjustment state of the (M-1) K + N th group of samples and the prediction result of the (M-1) K + N th group of samples into a preset reward function to obtain (M-1) K + N reward values; adjusting the first neural network model based on the (M-1) K + N reward values to obtain an adjusted (M-1) K + N first neural network model; under the condition that the (M-1) K + N is smaller than the preset accumulated step length, taking N = N +1 and the (M-1) K + N first neural network model as a first neural network model to be input, and repeating the steps until the (M-1) K + N is equal to the preset accumulated step length; correspondingly, in the case that N is equal to the preset accumulated step size, determining, from the N first neural network models, a first neural network model meeting the preset reward condition includes: and under the condition that the (M-1) K + N is equal to the preset accumulated step length, determining a first neural network model meeting a preset reward condition from the (M-1) K + N first neural network models.
Optionally, the step of inputting the device state into the state optimization model to obtain an optimized device state of the production line device further includes; acquiring adjustment data of the state optimization model; the adjustment data comprises the device state and the optimized device state; and displaying the adjustment data in a form of a chart.
Optionally, the reward function comprises a first correlation function and a second correlation function; the first correlation function is used for representing a prediction result corresponding to the sample adjustment state; the second correlation function is used to represent the degree of adjustment between the sample adjustment state and the sample device state.
Optionally, the obtaining a pre-trained state optimization model further includes: acquiring an original data set of a production line manufacturing process; the original data set comprises historical equipment states of the production line equipment and defect labels corresponding to the historical equipment states; carrying out data preprocessing on the original data set to obtain an edited data set; acquiring a preset second neural network model; and training the second neural network model based on the edited data set to obtain the defect prediction model.
Optionally, the performing data preprocessing on the original data set to obtain an edited data set includes: performing data supplement on the missing control instructions and/or process parameters in the original data set so as to eliminate the influence of the missing control instructions and/or process parameters on model training; carrying out data expansion on historical equipment states in the original data set so as to increase the number of the historical equipment states; and screening the historical equipment states in the original data set to reduce the data volume participating in model training.
Optionally, the step of repeating the training on the first neural network model meeting the preset reward condition until a preset convergence condition or a preset training number is reached, and using the obtained current first neural network model as a state optimization model includes: acquiring the current training times; and under the condition that the current training times are smaller than the preset training times, acquiring reward values in multiple training processes before the current training times to form a reward value set. Generating an incentive value curve based on the incentive value set, and judging whether the incentive value curve is converged; and under the condition that the reward value curve is converged, taking the obtained current first neural network model as a state optimization model.
In a second aspect, there is provided an automatic management device for production line defects, the device comprising:
a state acquisition module: the method comprises the steps of acquiring the equipment state corresponding to the production line equipment when the product manufactured by the production line has defects; the equipment state comprises control instructions and process parameters of the production line equipment;
a model acquisition module: the state optimization model is used for obtaining pre-training; the obtaining of the pre-trained state optimization model includes: inputting the equipment states of the Nth group of samples into the first neural network model to obtain an Nth group of sample adjustment states corresponding to the equipment states of the Nth group of samples; n is an integer greater than or equal to 1; the sample device states include K groups of sample device states, K being an integer greater than or equal to 1; inputting the adjustment state of the Nth group of samples into a pre-trained defect prediction model to obtain a prediction result of the Nth group of samples; inputting the device state of the Nth group of samples, the adjustment state of the Nth group of samples and the prediction result of the Nth group of samples into a preset reward function to obtain an Nth reward value; adjusting the first neural network model based on the Nth reward value to obtain an adjusted Nth first neural network model; inputting N = N +1, inputting the state of the N +1 group of sample devices into the Nth first neural network model, and repeating the steps until N is equal to a preset accumulative step length; under the condition that the N is equal to the preset accumulated step length, determining a first neural network model meeting a preset reward condition from N first neural network models; repeating the training steps on the first neural network model which meets the preset reward condition until a preset convergence condition or a preset training frequency is reached, and taking the obtained current first neural network model as a state optimization model;
a model input module: the state optimization model is used for inputting the equipment state into the state optimization model to obtain the optimized equipment state of the production line equipment;
an equipment optimization module: for inputting the optimized device status into the line device to optimize the manufacturing defects of the line device.
In a third aspect, an electronic device is provided, which includes a memory, a controller and a computer program stored in the memory and operable on the controller, and the controller implements the steps of the automatic management method for production line defects when executing the computer program.
In a fourth aspect, a computer-readable storage medium is provided, wherein the storage medium stores a program, and the program is used for realizing the automatic management method for production line defects provided by the first aspect when executed by a processor.
The beneficial effect of this application includes at least: acquiring the equipment state corresponding to the production line equipment when the product manufactured by the production line has defects; the equipment state comprises control instructions and process parameters of production line equipment; acquiring a pre-trained state optimization model; inputting the equipment state into a state optimization model to obtain an optimized equipment state of the production line equipment; inputting the optimized equipment state into the production line equipment to optimize the manufacturing defects of the production line equipment; obtaining a pre-trained state optimization model, comprising: inputting the device states of the Nth group of samples into the first neural network model to obtain an Nth group of sample adjustment states corresponding to the device states of the Nth group of samples; n is an integer greater than or equal to 1; the sample device states comprise K groups of sample device states, wherein K is an integer greater than or equal to 1; inputting the adjustment state of the Nth group of samples into a pre-trained defect prediction model to obtain a prediction result of the Nth group of samples; inputting the equipment state of the Nth group of samples, the adjustment state of the Nth group of samples and the prediction result of the Nth group of samples into a preset reward function to obtain an Nth reward value; adjusting the first neural network model based on the Nth reward value to obtain an adjusted Nth first neural network model; inputting the state of the N +1 group of sample devices into the Nth first neural network model, and repeating the steps until N is equal to the preset accumulated step length; under the condition that N is equal to the preset accumulated step length, determining a first neural network model meeting the preset reward condition from N first neural network models; and repeating the training steps on the first neural network model which meets the preset reward condition until the preset convergence condition or the preset training times is reached, and taking the obtained current first neural network model as a state optimization model.
The method can solve the problems that the defects of processing equipment in the production line are analyzed and detected in a manual mode, optimization is carried out, a large amount of time and energy are consumed, and the defect management efficiency of the production line is low. The state optimization model can directly output the corresponding optimized equipment state according to the input equipment state and input the optimized equipment state into the production line equipment so as to improve the control instruction and the process parameters of the production line equipment.
In addition, adjusting data of the state optimization model is obtained; the adjustment data comprises a device state and an optimized device state; the adjustment data is presented in the form of a graph. The adjustment condition of the equipment state can be visually seen.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments or the technical solutions in the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a schematic diagram of a method for automatic management of defects in a production line according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a second neural network model training process provided by one embodiment of the present application;
FIG. 3 is a schematic diagram of a first neural network model training process provided by one embodiment of the present application;
FIG. 4 is a block diagram of an apparatus for automatic management of production line defects according to an embodiment of the present application;
FIG. 5 is a block diagram of an electronic device provided by an embodiment of the application.
Detailed Description
The technical solutions of the present application will be described clearly and completely with reference to the accompanying drawings, and it is to be understood that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
In this application, where the context does not dictate to the contrary, the use of directional terms such as "upper, lower, top, bottom" generally refers to the orientation as shown in the drawings, or to the component itself in a vertical, perpendicular, or gravitational orientation; likewise, for ease of understanding and description, "inner and outer" refer to the inner and outer relative to the profile of the components themselves, but the above directional words are not intended to limit the application.
The following describes the method for automatically managing defects of a production line in detail.
As shown in fig. 1, an embodiment of the present application provides an automatic management method for production line defects, where implementation of the method may rely on a computer program, and the computer program may be run on a computer device such as a smart phone, a tablet computer, a personal computer, or the like, or run on a server, and the embodiment does not limit an operation subject of the method. The method at least comprises the following steps:
step 101, acquiring a device state corresponding to production line equipment when a product manufactured by a production line has a defect.
Wherein the equipment state comprises control instructions and process parameters of the production line equipment.
Such as: taking the production line as a semiconductor production line as an example, the production line equipment includes a wafer scriber, an oxidation furnace, a chemical vapor deposition system, a photolithography machine, an ion etching system, an ion implanter, and the like. Accordingly, the equipment status may include control commands and process parameters of the production line equipment during wafer dicing, contaminant removal, thermal oxidation, exposure, development, etc.
The state optimization model is used for optimizing the input equipment state and outputting an optimized equipment state, wherein the optimized equipment state comprises the optimized control instruction and process parameters of the production line equipment with product defects.
In this embodiment, the state optimization model is obtained by training based on a preset first neural network model. The first Neural network model may be a Deep Neural Network (DNN) or a Convolutional Neural Network (CNN), and the embodiment does not limit the type of the first Neural network model.
In this embodiment, the first neural network model is trained according to K groups of sample device states to obtain a state optimization model. Through setting K groups of sample equipment states, the data volume of model training can be increased, so that the model training result is more accurate. Wherein K is an integer greater than or equal to 1.
Specifically, obtaining a pre-trained state optimization model at least includes steps S11 to S15:
and S11, inputting the equipment states of the Nth group of samples into the first neural network model to obtain the adjustment states of the Nth group of samples corresponding to the equipment states of the Nth group of samples.
Wherein N is an integer greater than or equal to 1; the sample device states include K sets of sample device states.
In this embodiment, the state of the sample device includes historical control instructions and historical process parameters corresponding to the production line device when the product manufactured by the production line has a defect.
Accordingly, the sample adjustment state comprises historical control instructions and historical process parameter adjusted control instructions and process parameters.
Such as: taking the process parameters as the cutting angle and the heating temperature as examples, if the cutting angle in the input sample device state is 10 degrees and the heating temperature is 15 degrees, the cutting angle in the sample adjustment state output by the first neural network model may be 12 degrees, and the heating temperature may be 13 degrees.
In practical implementation, since both the control command and the process parameter in the production line equipment have an adjustment range, for example, taking the process parameter as the temperature, the adjustment range of the temperature may be 10 degrees to 100 degrees or 20 degrees to 100 degrees, and so on, the first neural network model also has an adjustment range for adjusting the state of the sample equipment.
Such as: taking the production line device as an example of the heating device, correspondingly, the process parameter includes a heating range of the heating device, where the heating range is 0 to 100 degrees celsius, and then the adjustment range of the first neural network model for the temperature of the heating device in the sample device state is also 0 to 100 degrees celsius.
After obtaining the sample adjustment state, since it is not possible to determine whether the sample adjustment state can reduce defects of products in the manufacturing process of the production line, it is necessary to perform defect prediction on the sample adjustment state.
In this embodiment, the defect prediction model is used to perform defect prediction on the sample adjustment state output by the first neural network model, and the sample adjustment state is input into the defect prediction model, so that the defect prediction model outputs a corresponding prediction result, where the prediction result is used to indicate whether a product produced by the production line equipment corresponding to the sample adjustment state has a defect.
Accordingly, the defect prediction model is used to predict the defect of the sample adjustment state output by the first neural network model, and the second neural network model is trained to obtain the defect prediction model.
The second neural network model may be a recurrent neural network or a convolutional neural network, and the embodiment does not limit the type of the second neural network model.
In particular, reference is made to the schematic diagram of the second neural network model training process shown in FIG. 2. Obtaining a pre-trained state optimization model, further comprising: acquiring an original data set of a production line manufacturing process; the original data set comprises historical equipment states of the production line equipment and defect labels corresponding to the historical equipment states; carrying out data preprocessing on the original data set to obtain an edited data set; acquiring a preset second neural network model; and training the second neural network model based on the edited data set to obtain a defect prediction model.
The historical equipment state comprises historical control instructions and historical process parameters of the production line equipment; the defect label is used for indicating the defect condition of the product when the production line equipment produces the product according to the historical equipment state corresponding to the defect label.
In addition, problems may exist in the original data set, such as incomplete data or the amount of data does not meet the training requirements. Therefore, there is a possibility that the original data set cannot be directly used for model training, and based on this, data preprocessing needs to be performed on the original data set to obtain an edited data set.
The following description of data preprocessing is given in three cases:
first, there is a data loss problem with the original data set.
During data collection, data in the original data set may be missing due to incomplete data collection in production line equipment or missing of collected data, and therefore, the original data set needs to be detected to determine a missing portion of data in the original data set and perform data supplementation on the missing portion.
The step of performing data supplementation on the missing part in the original data set refers to performing data supplementation on the missing control instruction and/or process parameter so as to eliminate the influence of the missing control instruction and/or process parameter on model training.
In this embodiment, a complementary algorithm may be used to complement the missing control command and the process parameter, where the complementary algorithm may be a weighted k-neighborhood (KNN), a mean interpolation, or a median interpolation, and a specific selection of the complementary algorithm is not limited herein.
Second, the number of historical device states in the raw data set is small.
In this embodiment, the smaller number of the historical device states means that the number of the device states having defects in the product produced by the production line device is smaller, and since the smaller number of the historical device states causes the defect prediction model to be over-fitted, in the case of the smaller number of the historical device states, the historical device states in the original data set need to be subjected to data expansion to increase the number of the historical device states.
The method for expanding the state of the sample equipment in the original data set comprises the following steps: a few oversampling techniques (Synthetic minimum Over-sampling Technique, SMOTE), a weighted k-neighborhood method, and a Decision Tree method (Decision Tree) are synthesized, and a specific choice of the expansion method is not limited herein.
Third, the number of historical device states in the raw data set is high.
In this embodiment, the greater number of the historical device states means that the number of the acquired control instructions and the process parameters is greater, and the greater number of the historical device states in the original data set may cause complex model training and may cause low training efficiency, so that the historical device states in the original data set need to be screened when the number of the historical device states in the original data set is greater, so as to reduce the amount of data participating in the model training.
The screening method for screening the historical equipment states in the original data set comprises the following steps: a Recursive feature elimination with Cross Validation (RFECV), a Random forest method (Random forest), and the like, and the specific selection of the screening method is not limited herein.
And after the edited data set is obtained, training a preset second neural network model based on the edited data set to obtain a defect prediction model. The specific training process of the second neural network model belongs to the existing technical scheme, and is not described herein again.
Step S12: inputting the adjustment state of the Nth group of samples into a pre-trained defect prediction model to obtain a prediction result of the Nth group of samples; and inputting the equipment state of the Nth group of samples, the adjustment state of the Nth group of samples and the prediction result of the Nth group of samples into a preset reward function to obtain an Nth reward value.
In this embodiment, the reward function includes a first correlation function and a second correlation function, and the first correlation function is used to determine a prediction result corresponding to the sample adjustment state; the second correlation function is used to determine the degree of adjustment between the sample adjustment state and the sample device state. Specifically, the reward function may be represented by the following equation:
r(x，y)＝z(x，y)+l dist (x，y)
where r (x, y) represents a reward function, z (x, y) represents a first correlation function, l dist (x, y) represents a second correlation function.
Wherein the first correlation function may be represented by:
z(x，y)＝z(l pred (x；h w )，l pred (y；h w ))
wherein z (x, y) represents a first correlation function, x represents a sample device state, y represents a sample adjustment state, h w Is a defect prediction model, l pred (x；h w ) The state of the sample equipment is input into a defect prediction model to obtain a prediction result l pred (y；h w ) And (4) indicating the sample adjustment state and inputting the sample adjustment state into a defect prediction model to obtain a prediction result.
In this example, by mixing pred (x；h w ) The obtained prediction result is compared with pred (y；h w ) The obtained prediction results are compared to determine the value of the first correlation function, and the values of the first correlation function corresponding to different comparison results are different.
At l pred (x；h w ) The obtained prediction result is compared with pred (y；h w ) Under the condition that the obtained prediction results are the same, namely under the condition that the prediction result of the sample adjustment state indicates that the product produced by the production line equipment corresponding to the sample adjustment state has defects, the value of the first correlation function can be-1; at l pred (x；h w ) The obtained prediction result is compared with pred (y；h w ) In a case that the obtained prediction results are different, that is, in a case that the prediction result of the sample adjustment state indicates that the product produced by the production line device corresponding to the sample adjustment state is defect-free, the value of the first correlation function may be 1; the embodiment does not limit the implementation manner of the value of the first correlation function.
In this embodiment, the second correlation function may be represented by the following formula:
l dist (x，y)＝∑(x-y) 2
in the formula I dist (x, y) represents the second correlation function, x represents the sample device state, and y represents the sample adjustment state.
In this embodiment, the reward function is used to obtain a sample adjustment state closest to the state of the sample device, that is, the reward function is set to train the first neural network model, so that the sample adjustment state output by the first neural network model according to the input state of the sample device can reduce defects of products manufactured by the production line and can be closest to the state of the sample device. The closest sample device state represents the least degree of adjustment of the sample adjustment state.
And the traditional model training mode is to set a loss function to train the first neural network model. For example, the loss function can be represented by:
in the formula (I), the compound is shown in the specification,
indicating the state of the sample deviceThe number, x, represents the sample device state,
representing the formed set of sample device states, i.e.
g θ Representing a first neural network model, h w Is a defect prediction model,/ pred Indicating that the corresponding equipment state or the sample adjustment state of the sample is input into the defect prediction model to obtain a prediction result l dist Indicating the degree of adjustment between the sample adjustment state and the sample apparatus state, W 1 And W 2 Is 1 pred And l dist Corresponding weight, W 1 And W 2 Is in the range of 0 to 1.
However, training the first neural network model through the loss function may lead to a more complicated model training process and a lower training efficiency.
In order to solve the above problem, in this embodiment, a first neural network model is trained through a reward function, taking fig. 3 as an example, the first neural network model outputs a corresponding sample adjustment state according to an input sample device state, inputs the sample adjustment state into a defect prediction model to obtain a corresponding prediction result, inputs the sample device state, the sample adjustment state and the corresponding prediction result into the reward function to obtain a corresponding reward value, and trains the first neural network model based on the reward value to obtain a state optimization model. Therefore, a complex gradient solving process through a loss function is avoided, and the training efficiency can be improved.
Step S13: and adjusting the first neural network model based on the Nth reward value to obtain an adjusted Nth first neural network model.
The adjustment of the first neural network model based on the nth reward value refers to the adjustment of model parameters of the first neural network model based on the reward value.
For example, if the heating temperature in the sample adjustment state corresponding to the 1 st bonus value is 15 degrees, adjusting the first neural network model based on the 1 st bonus value includes adjusting model parameters of the first neural network model to adjust the heating temperature in the sample adjustment state.
And S14, inputting the N = N +1, inputting the state of the N +1 group of sample devices into the Nth first neural network model, and repeating the steps until N is equal to the preset accumulation step length.
The N = N +1 is used for training the first neural network model through a plurality of sample device states in one training process, so as to avoid that training only using one sample device state in one training process may cause inaccurate model training.
Meanwhile, the number of the states of the sample equipment in one training process is determined by setting a preset accumulated step length, namely, when the number of the input states of the sample equipment reaches the preset accumulated step length, one training process is finished.
In this embodiment, the preset accumulated step size may be any integer from 1 to 10, or may be any integer from 10 to 100, and the value of the preset accumulated step size is not limited in this embodiment.
Such as: taking the preset accumulated step length of 10 and n of 9 as an example, the device state of the 9 th group of samples is input into the 8 th first neural network model, and the adjustment state of the 9 th group of samples corresponding to the device state of the 9 th group of samples is obtained. Inputting the adjustment state of the 9 th group of samples into a defect prediction model to obtain a 9 th group of sample prediction results; inputting the device state of the 9 th group of samples, the adjustment state of the 9 th group of samples and the prediction result of the 9 th group of samples into a reward function to obtain a 9 th reward value; and adjusting the first neural network model based on the 9 th reward value to obtain an adjusted 9 th first neural network model. And repeating the steps once more when the N = N +1 and the N is 10, so as to obtain a corresponding 10 th first neural network model.
In addition, since there may be a case where the number K of sample device states is less than the preset accumulation step, the same sample device state needs to be repeatedly used for model training.
Specifically, inputting N = N +1 and the N +1 th group of sample device states into the nth first neural network model, and repeating the above steps until N is equal to a preset cumulative step size, including: when N is equal to K and K is smaller than a preset accumulated step length, acquiring a variable M, wherein the variable M is used for recording the times of N = 1; taking the Nth first neural network model as a first neural network model to be input; inputting the N =1, M = M +1, and the N group of sample device states into a first neural network model to be input, so as to obtain an (M-1) K + N group of sample adjustment states corresponding to the N group of sample device states; inputting the adjustment states of the (M-1) K + N groups of samples into a pre-trained defect prediction model to obtain the prediction results of the (M-1) K + N groups of samples; inputting the device state of the Nth group of samples, the adjustment state of the (M-1) K + N groups of samples and the prediction result of the (M-1) K + N groups of samples into a preset reward function to obtain (M-1) K + N reward values; adjusting the first neural network model based on the (M-1) K + N reward values to obtain an adjusted (M-1) K + N first neural network model; under the condition that (M-1) K + N is smaller than a preset accumulation step length, taking N = N +1 and taking the (M-1) K + N first neural network models as first neural network models to be input, and repeating the steps until (M-1) K + N is equal to the preset accumulation step length; correspondingly, under the condition that N is equal to the preset accumulated step length, determining a first neural network model meeting the preset reward condition from the N first neural network models, including: and under the condition that the (M-1) K + N is equal to the preset accumulated step length, determining a first neural network model meeting the preset reward condition from the (M-1) K + N first neural network models.
Such as: if K is 4, i.e., the sample device status includes 4 sets of sample device statuses, the preset accumulation step size is 10. When N =4+1, since there is no 5 th group of sample device states for input into the 4 th first neural network model, it is necessary to obtain the variable M to record the number of times the sample device state is reused, i.e., the number of times N = 1. Resetting N to 1, wherein the variable M is 2 at the moment, inputting the 1 st group of sample equipment states into the 4 th first neural network model to obtain the (M-1) K + N first neural network models corresponding to the 1 st group of sample equipment states, namely the 5 th group of sample adjustment states. When N reaches 4+1 again, resetting N to 1 again, wherein the variable M is 3, inputting the 1 st group of sample equipment states into the 8 th first neural network model to obtain the (M-1) th K + N group of sample adjustment states corresponding to the 1 st group of sample equipment states, namely, the 9 th group of sample adjustment states, and repeating the steps.
And S15, under the condition that N is equal to the preset accumulated step length, determining a first neural network model meeting the preset reward condition from the N first neural network models.
The preset reward condition is that a first neural network model which enables the prediction result of the sample adjustment state to be different from the result corresponding to the sample equipment state is selected, namely the prediction result of the sample adjustment state indicates that products produced by the production line equipment corresponding to the sample adjustment state are not defective, and the adjustment degree between the sample adjustment state and the sample equipment state is minimum.
In this embodiment, the first neural network model meeting the predetermined reward condition can be represented by the following formula:
in the formula, theta * Representing a first neural network model meeting preset reward conditions, r (x, y) representing a reward function, and T being a preset accumulated step length.
Such as: assuming that the preset cumulative step length is 10, when N is equal to 10, selecting the first neural network model corresponding to the maximum reward value of the 10 reward values according to a preset reward condition from the 10 first neural network models, and using the first neural network model as the first neural network model obtained in the training process for the next training process. Meanwhile, the first neural network model needs to meet the requirement that the obtained prediction result of the sample adjustment state is different from the result corresponding to the sample device state, that is, the product produced by the production line device corresponding to the sample adjustment state obtained through the adjustment of the first neural network model has no defect. If none of the 10 first neural network models obtained in the training process can meet the requirement that the obtained prediction result of the sample adjustment state is different from the result corresponding to the sample device state, the requirement can be skipped, only the first neural network model corresponding to the maximum reward value is selected, or the preset accumulated step length can be enlarged, and the processing means under the condition is not limited.
In order to determine a termination node of the training of the model, the termination of the training of the model is controlled by setting a training termination condition of the training of the model, wherein the training termination condition comprises a preset convergence condition and a preset training time.
Specifically, the training steps are repeated for the first neural network model meeting the preset reward condition until a preset convergence condition or a preset training frequency is reached, and the obtained current first neural network model is used as a state optimization model.
The preset convergence condition means that the reward value tends to be stable in multiple training processes.
In one example, the termination condition includes a preset convergence condition and a preset number of training times.
Under the condition that the termination condition comprises a preset convergence condition and a preset training frequency, when the first neural network model meets any one of the preset convergence condition and the preset training frequency, terminating the training of the first neural network model, and taking the obtained current first neural network model as a state optimization model.
Specifically, the above training steps are repeated for the first neural network model that meets the preset reward condition until a preset convergence condition or a preset training number is reached, and the obtained current first neural network model is used as a state optimization model, including: acquiring the current training times; and under the condition that the current training times are less than the preset training times, acquiring the reward value in the multiple training processes before the current training times to form a reward value set. Generating an incentive value curve based on the incentive value set, and judging whether the incentive value curve is converged; and under the condition that the reward value curve is converged, taking the obtained current first neural network model as a state optimization model.
The reward value set is used for generating a reward value curve so as to judge whether the first neural network model converges or not. The multiple training process may be two training processes or ten training processes, and the number of the multiple training processes is not limited herein.
Such as: the preset training times are 5000 times, when 3000 times are reached, a reward value set is formed by obtaining reward values between 2950 th time and 3000 th time, after a reward value curve is generated, it is determined that the first neural network model is converged, model training is finished, and the first neural network model obtained in the 3000 th training process is output as a state optimization model.
In another example, the termination condition includes any one of a preset convergence condition and a preset training time, and when the preset convergence condition or the preset training time is reached, the model training is stopped to be terminated, and the first neural network model obtained in the last training process is output as the state optimization model.
And 103, inputting the equipment state into the state optimization model to obtain the optimized equipment state of the production line equipment.
In this embodiment, in order to enable a user to visually see a change of the device state, the device state input by the state optimization model and the optimized device state output by the state optimization model need to be obtained, and the device state and the optimized device state are visualized, where the visualization includes displaying adjustment data corresponding to the adjustment of the device state to the optimized device state in a form of a chart, and labeling the adjustment data in the chart. Wherein, the chart can be a table, a line chart, a tree chart and the like,
such as: if the device state includes a cutting angle of 15 degrees, the heating temperature is 20 degrees. And if the corresponding optimized equipment state comprises that the cutting angle is 14 degrees and the heating temperature is 25 degrees, adjusting the data to be the data corresponding to the cutting angle and the heating temperature, and highlighting or annotating the cutting angle and the heating temperature in the optimized equipment state.
Specifically, after the device state is input into the state optimization model and the optimized device state of the production line device is obtained, the method further includes: acquiring adjustment data of the state optimization model; the adjustment data includes a device state and an optimized device state; the adjustment data is presented in the form of a graph.
And 104, inputting the optimized equipment state into the production line equipment to optimize the manufacturing defects of the production line equipment.
In this embodiment, the step of inputting the optimized device status into the production line device includes replacing the optimized device status with the device status in the production line device, so that the production line device operates according to the optimized device status, so as to improve the defect condition of the production line device in producing products.
Such as: taking a wafer dicing saw in a semiconductor production line as an example, the dicing depth in the device state is 1.5 μm, and the dicing depth in the corresponding optimized device state output by the state optimization model is 1.4 μm, 1.4 μm is substituted for the original 1.5 μm.
In summary, the method for automatically managing defects in a production line provided by the embodiment obtains the device status corresponding to the production line device when the product manufactured by the production line has defects; the equipment state comprises control instructions and process parameters of production line equipment; acquiring a pre-trained state optimization model; inputting the equipment state into a state optimization model to obtain an optimized equipment state of the production line equipment; inputting the optimized equipment state into the production line equipment to optimize the manufacturing defects of the production line equipment; obtaining a pre-trained state optimization model, comprising: inputting the device states of the Nth group of samples into the first neural network model to obtain an Nth group of sample adjustment states corresponding to the device states of the Nth group of samples; n is an integer greater than or equal to 1; the sample device states comprise K groups of sample device states, wherein K is an integer greater than or equal to 1; inputting the adjustment state of the Nth group of samples into a pre-trained defect prediction model to obtain a prediction result of the Nth group of samples; inputting the Nth group of sample equipment states, the Nth group of sample adjustment states and the Nth group of sample prediction results into a preset reward function to obtain an Nth reward value; adjusting the first neural network model based on the Nth reward value to obtain an adjusted Nth first neural network model; inputting the state of the N +1 group of sample devices into the Nth first neural network model, and repeating the steps until N is equal to the preset accumulated step length; under the condition that N is equal to the preset accumulated step length, determining a first neural network model meeting the preset reward condition from N first neural network models; and repeating the training steps on the first neural network model which meets the preset reward condition until reaching the preset convergence condition or the preset training times, and taking the obtained current first neural network model as a state optimization model.
The method can solve the problems that the defects of processing equipment in a production line are analyzed and detected in a manual mode, optimization is carried out, a large amount of time and energy are consumed, and the defect management efficiency of the production line is low. The state optimization model can directly output the corresponding optimized equipment state according to the input equipment state and input the optimized equipment state into the production line equipment so as to improve the control instruction and the process parameters of the production line equipment.
In addition, adjusting data of the state optimization model is obtained; the adjustment data comprises a device state and an optimized device state; the adjustment data is presented in the form of a graph. The adjustment condition of the equipment state can be visually seen.
The present embodiment provides an automatic management device for production line defects, as shown in fig. 4. The device comprises at least the following modules: a state acquisition module 410, a model acquisition module 420, a model input module 430, and a device optimization module 440.
The state acquiring module 410 is used for acquiring the equipment state corresponding to the production line equipment when the product manufactured by the production line has defects; the equipment state includes control instructions and process parameters of the production line equipment.
A model obtaining module 420, configured to obtain a pre-trained state optimization model; obtaining a pre-trained state optimization model, comprising: inputting the equipment states of the Nth group of samples into the first neural network model to obtain an Nth group of sample adjustment states corresponding to the equipment states of the Nth group of samples; n is an integer greater than or equal to 1; the sample device states include K sets of sample device states, K being an integer greater than or equal to 1; inputting the adjustment state of the Nth group of samples into a pre-trained defect prediction model to obtain a prediction result of the Nth group of samples; inputting the Nth group of sample equipment states, the Nth group of sample adjustment states and the Nth group of sample prediction results into a preset reward function to obtain an Nth reward value; adjusting the first neural network model based on the Nth reward value to obtain an adjusted Nth first neural network model; inputting the state of the N +1 group of sample devices into the Nth first neural network model, and repeating the steps until N is equal to the preset accumulated step length; under the condition that N is equal to a preset accumulated step length, determining a first neural network model meeting a preset reward condition from N first neural network models; and repeating the training steps on the first neural network model which meets the preset reward condition until the preset convergence condition or the preset training times is reached, and taking the obtained current first neural network model as a state optimization model.
And the model input module 430 is configured to input the device state into the state optimization model to obtain an optimized device state of the production line device.
The device optimization module 440 is configured to input the optimized device status into the production line device to optimize the manufacturing defects of the production line device.
Reference is made to the above method and apparatus embodiments for relevant details.
It should be noted that: in the above embodiment, when performing the automatic management of the production line defect, the automatic management device for the production line defect is only illustrated by the division of the functional modules, and in practical applications, the above function distribution may be completed by different functional modules according to needs, that is, the internal structure of the automatic management device for the production line defect is divided into different functional modules to complete all or part of the above described functions. In addition, the embodiment of the automatic management device for production line defects and the embodiment of the automatic management method for production line defects provided by the above embodiments belong to the same concept, and the specific implementation process is described in the embodiment of the method for details, which is not described herein again.
The present embodiment provides an electronic device as shown in fig. 5. The electronic device includes at least a processor 510 and a memory 520.
Processor 510 may include one or more processing cores such as: 4 core processors, 8 core processors, etc. The processor 510 may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). Processor 510 may also include a main processor and a coprocessor, where the main processor is a processor, also called a Central Processing Unit (CPU), for Processing data in the wake-up state; a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 510 may be integrated with a GPU (Graphics Processing Unit) that is responsible for rendering and drawing the content that the display screen needs to display. In some embodiments, processor 510 may also include an AI (Artificial Intelligence) processor for processing computational operations related to machine learning.
Memory 520 may include one or more computer-readable storage media, which may be non-transitory. Memory 520 may also include high speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer readable storage medium in the memory 520 is configured to store at least one instruction for execution by the processor 510 to implement the automatic management method for defects in a production line provided by the method embodiments of the present application.
In some embodiments, the electronic device may further include: a peripheral interface and at least one peripheral. The processor 510, memory 520, and peripheral interface may be connected by buses or signal lines. Each peripheral may be connected to the peripheral interface via a bus, signal line, or circuit board. Illustratively, peripheral devices include, but are not limited to: radio frequency circuit, touch display screen, audio circuit, power supply, etc.
Of course, the electronic device may include fewer or more components, which is not limited by the embodiment.
Optionally, the present application further provides a computer-readable storage medium, in which a program is stored, and the program is loaded and executed by a processor to implement the automatic management method for production line defects of the foregoing method embodiments.
All possible combinations of the technical features in the above embodiments may not be described for the sake of brevity, but should be considered as being within the scope of the present disclosure as long as there is no contradiction between the combinations of the technical features.
It is to be understood that the above-described embodiments are only a few, but not all, of the embodiments described herein. Based on the embodiments in the present application, a person skilled in the art may make other variations or changes without creative efforts, and all of them should fall into the protection scope of the present application.
Claims (10)
1. A method for automatic management of production line defects, the method comprising:
acquiring the equipment state corresponding to the production line equipment when the product manufactured by the production line has defects; the equipment state comprises control instructions and process parameters of the production line equipment;
acquiring a pre-trained state optimization model;
inputting the equipment state into the state optimization model to obtain an optimized equipment state of the production line equipment;
inputting the optimized equipment state into the production line equipment to optimize the manufacturing defects of the production line equipment;
the obtaining of the pre-trained state optimization model includes:
inputting the equipment states of the Nth group of samples into a first neural network model to obtain an Nth group of sample adjustment states corresponding to the equipment states of the Nth group of samples; n is an integer greater than or equal to 1; the sample device states include K sets of sample device states, K being an integer greater than or equal to 1;
inputting the adjustment state of the Nth group of samples into a pre-trained defect prediction model to obtain a prediction result of the Nth group of samples;
inputting the device state of the Nth group of samples, the adjustment state of the Nth group of samples and the prediction result of the Nth group of samples into a preset reward function to obtain an Nth reward value;
adjusting the first neural network model based on the Nth reward value to obtain an adjusted Nth first neural network model;
inputting N = N +1, inputting the state of the N +1 group of sample devices into the Nth first neural network model, and repeating the steps until N is equal to a preset accumulative step length;
under the condition that the N is equal to the preset accumulated step length, determining a first neural network model meeting a preset reward condition from N first neural network models;
and repeating the training steps on the first neural network model which meets the preset reward condition until a preset convergence condition or a preset training frequency is reached, and taking the obtained current first neural network model as a state optimization model.
2. The method of claim 1, wherein inputting N = N +1 and the N +1 set of sample device states into the nth first neural network model, and repeating the above steps until N equals a preset cumulative step size, comprises:
when N is equal to K and K is smaller than a preset accumulated step length, acquiring a variable M, wherein the variable M is used for recording the times of N = 1;
taking the Nth first neural network model as a first neural network model to be input;
inputting the N =1, M = M +1, and the Nth group of sample device states into the first neural network model to be input, so as to obtain an (M-1) K + N group of sample adjustment states corresponding to the Nth group of sample device states;
inputting the (M-1) K + N groups of sample adjustment states into a pre-trained defect prediction model to obtain an (M-1) K + N group of sample prediction results;
inputting the device state of the Nth group of samples, the adjustment state of the (M-1) K + N th group of samples and the prediction result of the (M-1) K + N th group of samples into a preset reward function to obtain (M-1) K + N reward values;
adjusting the first neural network model based on the (M-1) K + N reward values to obtain an (M-1) K + N adjusted first neural network model;
under the condition that the (M-1) K + N is smaller than the preset accumulated step length, taking N = N +1 and the (M-1) K + N first neural network model as a first neural network model to be input, and repeating the steps until the (M-1) K + N is equal to the preset accumulated step length;
correspondingly, in the case that N is equal to the preset accumulated step length, determining, from the N first neural network models, a first neural network model meeting the preset reward condition includes:
and under the condition that the (M-1) K + N is equal to the preset accumulated step length, determining a first neural network model meeting a preset reward condition from the (M-1) K + N first neural network models.
3. The method of claim 1, wherein said inputting said equipment state into said state optimization model further comprises, after obtaining an optimized equipment state for said production line equipment;
acquiring adjustment data of the state optimization model; the adjustment data comprises the device state and the optimized device state;
and displaying the adjusting data in a form of a chart.
4. The method of claim 1, wherein the reward function comprises a first correlation function and a second correlation function; the first correlation function is used for representing a prediction result corresponding to the sample adjustment state; the second correlation function is used to represent the degree of adjustment between the sample adjustment state and the sample device state.
5. The method of claim 1, wherein obtaining the pre-trained state optimization model further comprises:
acquiring an original data set of a production line manufacturing process; the original data set comprises historical equipment states of production line equipment and defect labels corresponding to the historical equipment states;
carrying out data preprocessing on the original data set to obtain an edited data set;
acquiring a preset second neural network model;
and training the second neural network model based on the edited data set to obtain the defect prediction model.
6. The method of claim 5, wherein the pre-processing the original data set to obtain an edited data set comprises:
performing data supplement on the missing control instructions and/or process parameters in the original data set so as to eliminate the influence of the missing control instructions and/or process parameters on model training;
carrying out data expansion on historical equipment states in the original data set so as to increase the number of the historical equipment states;
and screening the historical equipment states in the original data set to reduce the data volume participating in model training.
7. The method according to claim 1, wherein the step of repeating the training on the first neural network model meeting the predetermined reward condition until a predetermined convergence condition or a predetermined number of training times is reached, and using the obtained current first neural network model as a state optimization model comprises:
acquiring the current training times;
and under the condition that the current training times are less than the preset training times, acquiring the reward value in a plurality of training processes before the current training times to form a reward value set.
Generating an incentive value curve based on the incentive value set, and judging whether the incentive value curve is converged;
and under the condition that the reward value curve is converged, taking the obtained current first neural network model as a state optimization model.
8. An automatic management device for production line defects, characterized in that the device comprises:
a state acquisition module: the method comprises the steps of acquiring the equipment state corresponding to the production line equipment when the product manufactured by the production line has defects; the equipment state comprises control instructions and process parameters of the production line equipment;
a model acquisition module: the state optimization model is used for obtaining a pre-trained state optimization model; the obtaining of the pre-trained state optimization model includes:
inputting the device states of the Nth group of samples into the first neural network model to obtain the adjustment states of the Nth group of samples corresponding to the device states of the Nth group of samples; n is an integer greater than or equal to 1; the sample device states include K groups of sample device states, K being an integer greater than or equal to 1;
inputting the adjustment state of the Nth group of samples into a pre-trained defect prediction model to obtain a prediction result of the Nth group of samples;
inputting the device state of the Nth group of samples, the adjustment state of the Nth group of samples and the prediction result of the Nth group of samples into a preset reward function to obtain an Nth reward value;
adjusting the first neural network model based on the Nth reward value to obtain an adjusted Nth first neural network model;
inputting N = N +1, inputting the state of the N +1 group of sample devices into the Nth first neural network model, and repeating the steps until N is equal to a preset accumulative step length;
under the condition that the N is equal to the preset accumulated step length, determining a first neural network model meeting a preset reward condition from N first neural network models;
repeating the training steps on the first neural network model which meets the preset reward condition until a preset convergence condition or preset training times is reached, and taking the obtained current first neural network model as a state optimization model;
a model input module: the state optimization model is used for inputting the equipment state into the state optimization model to obtain the optimized equipment state of the production line equipment;
an equipment optimization module: for inputting the optimized device status into the line device to optimize the manufacturing defects of the line device.
9. An electronic device, characterized in that the electronic device comprises: a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the electronic device to perform the steps of the automatic management method of production line defects according to any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon instructions, when executed by a processor, for implementing the steps of the automatic management method of production line defects according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211368271.3A CN115640933B (en) | 2022-11-03 | 2022-11-03 | Automatic defect management method, device and equipment for production line and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211368271.3A CN115640933B (en) | 2022-11-03 | 2022-11-03 | Automatic defect management method, device and equipment for production line and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115640933A true CN115640933A (en) | 2023-01-24 |
| CN115640933B CN115640933B (en) | 2023-10-13 |
Family
ID=84947281
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211368271.3A Active CN115640933B (en) | 2022-11-03 | 2022-11-03 | Automatic defect management method, device and equipment for production line and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115640933B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117689086A (en) * | 2024-02-02 | 2024-03-12 | 山东国泰民安玻璃科技有限公司 | Production parameter optimization method, equipment and medium for medium borosilicate glass bottle |
| CN117908496A (en) * | 2024-03-13 | 2024-04-19 | 山东国泰民安玻璃科技有限公司 | Automatic production regulation and control method and system for tube-made glass bottles |
| CN118468614A (en) * | 2024-07-12 | 2024-08-09 | 江西安达电子有限公司 | Circuit board manufacturing process optimization method and equipment |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108897925A (en) * | 2018-06-11 | 2018-11-27 | 华中科技大学 | A kind of casting technological parameter optimization method based on casting defect prediction model |
| US20200327450A1 (en) * | 2019-04-15 | 2020-10-15 | Apple Inc. | Addressing a loss-metric mismatch with adaptive loss alignment |
| CN112416783A (en) * | 2020-11-25 | 2021-02-26 | 武汉联影医疗科技有限公司 | Method, device, equipment and storage medium for determining software quality influence factors |
| CN113239639A (en) * | 2021-06-29 | 2021-08-10 | 暨南大学 | Policy information generation method, policy information generation device, electronic device, and storage medium |
| WO2021169218A1 (en) * | 2020-02-26 | 2021-09-02 | 平安科技（深圳）有限公司 | Data pushing method and system, electronic device and storage medium |
| US20220067622A1 (en) * | 2020-08-25 | 2022-03-03 | Noodle Analytics, Inc. | Systems and methods for automating production intelligence across value streams using interconnected machine-learning models |
| US20220172070A1 (en) * | 2020-11-27 | 2022-06-02 | Shanghai Huali Integrated Circuit Corporation | Method and apparatus for predicting yield of semiconductor devices |
| CN114662676A (en) * | 2022-04-11 | 2022-06-24 | 深圳Tcl新技术有限公司 | Model optimization method and device, electronic equipment and computer-readable storage medium |
| CN114862832A (en) * | 2022-05-31 | 2022-08-05 | 北京中祥英科技有限公司 | Method, device and equipment for optimizing defect detection model and storage medium |
-
2022
- 2022-11-03 CN CN202211368271.3A patent/CN115640933B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108897925A (en) * | 2018-06-11 | 2018-11-27 | 华中科技大学 | A kind of casting technological parameter optimization method based on casting defect prediction model |
| US20200327450A1 (en) * | 2019-04-15 | 2020-10-15 | Apple Inc. | Addressing a loss-metric mismatch with adaptive loss alignment |
| WO2021169218A1 (en) * | 2020-02-26 | 2021-09-02 | 平安科技（深圳）有限公司 | Data pushing method and system, electronic device and storage medium |
| US20220067622A1 (en) * | 2020-08-25 | 2022-03-03 | Noodle Analytics, Inc. | Systems and methods for automating production intelligence across value streams using interconnected machine-learning models |
| CN112416783A (en) * | 2020-11-25 | 2021-02-26 | 武汉联影医疗科技有限公司 | Method, device, equipment and storage medium for determining software quality influence factors |
| US20220172070A1 (en) * | 2020-11-27 | 2022-06-02 | Shanghai Huali Integrated Circuit Corporation | Method and apparatus for predicting yield of semiconductor devices |
| CN113239639A (en) * | 2021-06-29 | 2021-08-10 | 暨南大学 | Policy information generation method, policy information generation device, electronic device, and storage medium |
| CN114662676A (en) * | 2022-04-11 | 2022-06-24 | 深圳Tcl新技术有限公司 | Model optimization method and device, electronic equipment and computer-readable storage medium |
| CN114862832A (en) * | 2022-05-31 | 2022-08-05 | 北京中祥英科技有限公司 | Method, device and equipment for optimizing defect detection model and storage medium |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117689086A (en) * | 2024-02-02 | 2024-03-12 | 山东国泰民安玻璃科技有限公司 | Production parameter optimization method, equipment and medium for medium borosilicate glass bottle |
| CN117689086B (en) * | 2024-02-02 | 2024-04-26 | 山东国泰民安玻璃科技有限公司 | Production parameter optimization method, equipment and medium for medium borosilicate glass bottle |
| CN117908496A (en) * | 2024-03-13 | 2024-04-19 | 山东国泰民安玻璃科技有限公司 | Automatic production regulation and control method and system for tube-made glass bottles |
| CN117908496B (en) * | 2024-03-13 | 2024-05-24 | 山东国泰民安玻璃科技有限公司 | Automatic production regulation and control method and system for tube-made glass bottles |
| CN118468614A (en) * | 2024-07-12 | 2024-08-09 | 江西安达电子有限公司 | Circuit board manufacturing process optimization method and equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115640933B (en) | 2023-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115640933A (en) | Method, device and equipment for automatically managing production line defects and storage medium | |
| KR102611938B1 (en) | Generate integrated circuit floorplans using neural networks | |
| KR102258942B1 (en) | System and method for the automatic determination of critical parametric electrical test parameters for inline yield monitoring | |
| CN115308562A (en) | Chip testing method and related equipment | |
| CN114444986A (en) | Product analysis method, system, device and medium | |
| US10338547B2 (en) | System, method and apparatuses for determining parameter settings for a power generation system and a tangible computer readable medium | |
| CN116339145A (en) | Machine parameter regulation and control method, equipment and storage medium | |
| CN115427968A (en) | Robust artificial intelligence reasoning in edge computing devices | |
| CN114757512A (en) | Job shop scheduling method based on graph neural network | |
| WO2021066194A1 (en) | Learning device, learning method, and learning program | |
| CN117520741A (en) | Method for predicting and improving yield of semiconductor factory based on big data | |
| CN114757111B (en) | Intelligent equipment health management method based on digital twinning technology | |
| CN114503244B (en) | Tool control using multi-stage LSTM for predicting measurements on wafers | |
| CN115169832A (en) | Sensitivity analysis method and system based on curve form change | |
| CN109766181A (en) | A kind of RMS schedulability determination method and device based on deep learning | |
| CN115018150A (en) | Sample generation method, model training method, storage medium, and program product | |
| CN113626420A (en) | Data preprocessing method and device and readable storage medium | |
| CN116542504B (en) | Parameter-adaptive semiconductor workpiece production scheduling method, equipment and storage medium | |
| CN111178780A (en) | Operation and maintenance strategy setting method and device, storage medium and electronic equipment | |
| CN117075684B (en) | Self-adaptive clock gridding calibration method for Chiplet chip | |
| WO2020251037A1 (en) | Learning device, extraction device, learning method, extraction method, learning program, and extraction program | |
| CN116992760A (en) | Method and system for predicting and explaining film thickness process of wafer deposition | |
| CN116068962B (en) | Process route planning method based on numerical control machine tool and related equipment | |
| CN118377210B (en) | Comprehensive processing method for combining quartile with ARIMA of atomic clock sequence | |
| JPH11150047A (en) | Semiconductor device manufacturing method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |