Equipment parameter prediction method and device based on machine learning, equipment and medium

Abstract

The invention discloses an equipment parameter prediction method and device based on machine learning, equipment and a medium, and relates to the technical field of machine learning and intelligent manufacturing, and the method comprises the steps: collecting initial operation data corresponding to a printing machine, a chip mounter and reflow soldering in real time, and storing the initial operation data to a preset database; the initial operation data comprises respective core parameters and yield information; preprocessing each initial operation data in a preset database to obtain target operation data, and processing each target operation data by using a plurality of machine learning models to obtain an initial parameter prediction result; the initial parameter prediction result comprises a first pass yield prediction result, a theoretical time consumption prediction result and a printing speed prediction result; and calling a preset reinforcement learning model, and processing the initial parameter prediction result by using a depth deterministic strategy gradient algorithm and a near-end strategy optimization algorithm to obtain a corresponding target parameter prediction result. Therefore, the quality of the obtained process parameters can be improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of machine learning and intelligent manufacturing, and particularly relates to a method, device, equipment and medium for predicting equipment parameters based on machine learning. Background Art

[0002] With the intensification of market competition, manufacturing enterprises are facing various challenges, including improving production efficiency, reducing production costs, and enhancing product quality. In the traditional manufacturing mode, the setting of process parameters of equipment often lacks a scientific basis, resulting in unstable product quality. This situation restricts the sustainable development of enterprises and also hinders the process of intelligent manufacturing to a certain extent.

[0003] Taking the surface mount technology (SMT) as an example, this technology is widely applied to equipment such as printers, mounters, and reflow soldering machines. In this process, the operating states and process parameters of various equipment play a decisive role in the quality of the final product. However, traditional methods often rely solely on the experience of operators and lack data-driven scientific decision-making. This current situation not only causes waste of resources but also increases the uncertainty in the production process, leading to a decrease in the product qualification rate.

[0004] To address the above challenges, more and more enterprises have begun to attempt to apply machine learning algorithms to the analysis and prediction of production data. As a powerful data analysis tool, machine learning can identify complex non-linear relationships through the learning of historical data, thereby realizing the intelligent prediction of process parameters. However, the current application of machine learning algorithms is still in its infancy, and there is a lack of systematic analysis of the relationships between different equipment and different process parameters, making the construction and application of machine learning models face many challenges.

[0005] As can be seen from the above, how to improve the quality of obtaining process parameters in the process of predicting equipment parameters based on machine learning is an urgent problem to be solved at present. Summary of the Invention

[0006] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for predicting equipment parameters based on machine learning, which can improve the quality and speed of obtaining process parameters, and further enhance the efficiency of the production process. The specific solutions are as follows:

[0007] In the first aspect, the present application provides a method for predicting equipment parameters based on machine learning, including:

[0008] Using a preset sensor and data acquisition device to collect the initial operation data corresponding to the printer, mounter, and reflow soldering machine in real time, and storing the initial operation data in a preset database; the initial operation data includes the corresponding core parameters and production information;

[0009] Preprocess each of the initial operation data in the preset database to obtain target operation data, and process each of the target operation data using a number of machine learning models to obtain an initial parameter prediction result; the initial parameter prediction result includes a throughput prediction result, a theoretical time consumption prediction result, and a printing speed prediction result;

[0010] Call a preset reinforcement learning model and use the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to process the initial parameter prediction result to obtain a corresponding target parameter prediction result, so as to configure the device using the target parameter prediction result.

[0011] Optionally, the real-time collection of the initial operation data corresponding to the printer, mounter, and reflow oven using a preset sensor and data acquisition device includes:

[0012] Use a preset sensor to collect the printing speed, demolding time, printing pressure, and printing temperature corresponding to the printer in real time;

[0013] Use a data acquisition device to collect the placement speed, placement accuracy, placement production volume, mounter operation time, and mounter fault information corresponding to the mounter in real time;

[0014] Use a preset sensor to collect the welding temperature curve, welding time, and welding furnace atmosphere corresponding to the reflow oven in real time.

[0015] Optionally, the preprocessing operation of each of the initial operation data in the preset database to obtain target operation data includes:

[0016] Perform operations to remove missing values, outliers, and duplicate data on each of the initial operation data in the preset database to obtain cleaned operation data;

[0017] Perform standardization processing on each of the cleaned operation data using a preset standardization operation to obtain standardized data;

[0018] Perform a combination and transformation operation on each of the standardized data using a preset row combination and transformation operation to obtain data to be extracted, and perform a feature extraction operation on the data to be extracted using a preset feature extraction method to obtain target operation data.

[0019] Optionally, the processing of each of the target operation data using a number of machine learning models to obtain an initial parameter prediction result includes:

[0020] Divide each of the target operation data according to a preset division rule to obtain a training set and a test set, and then train an initial decision tree regression model based on the training set and the test set to obtain a decision tree regression model to be evaluated;

[0021] Use a preset cross-validation method and preset model metrics to determine whether the decision tree regression model to be evaluated meets the preset performance conditions. If the decision tree regression model to be evaluated meets the preset performance conditions, set the decision tree regression model to be evaluated as the target decision tree regression model, and use the target decision tree regression model to process each piece of target operation data to obtain the corresponding throughput rate prediction result.

[0022] Optionally, the processing of each piece of target operation data by using a plurality of machine learning models to obtain an initial parameter prediction result includes:

[0023] Use the least squares method and train each initial time-consuming prediction formula in the initial linear regression model based on the core parameters corresponding to the printing process, the chip mounting process, and the soldering process respectively to obtain the target time-consuming prediction formulas corresponding to the printing process, the chip mounting process, and the soldering process respectively;

[0024] Use each target time-consuming prediction formula to perform a theoretical time-consuming calculation operation on the corresponding core parameters to obtain the corresponding theoretical time-consuming prediction results.

[0025] Optionally, the processing of each piece of target operation data by using a plurality of machine learning models to obtain an initial parameter prediction result includes:

[0026] Set a target K value based on user requirements, and use the K-nearest neighbor regression algorithm and based on the target K value and the front and rear squeegee pressures corresponding to the printing machine to train the initial printing speed prediction model to obtain a target printing speed prediction model;

[0027] Use the target printing speed prediction model to predict the printing speed of the printing machine to obtain a printing speed prediction result.

[0028] Optionally, the calling of a preset reinforcement learning model and using the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to process the initial parameter prediction result to obtain the corresponding target parameter prediction result includes:

[0029] Call a preset reinforcement learning model and use the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to process the state to be processed to obtain the corresponding action to be executed; the state to be processed includes the throughput rate prediction result, the theoretical time-consuming prediction result, and the printing speed prediction result;

[0030] Adjust the process parameters corresponding to the printing machine, the chip mounter, and the reflow soldering based on each of the to-be-executed actions to obtain a reward value corresponding to the to-be-executed action; the process parameters include printing speed, chip mounting accuracy, soldering temperature, and equipment operation rate;

[0031] Use a preset weighted comprehensive scoring rule and the Pareto optimal solution method and adjust the parameters of the preset reinforcement learning model based on the reward value, so as to process the initial parameter prediction result using the obtained adjusted preset reinforcement learning model to obtain a corresponding target parameter prediction result.

[0032] In a second aspect, the present application provides a device parameter prediction device based on machine learning, including:

[0033] An operation data acquisition module, configured to use a preset sensor and data acquisition device to collect initial operation data corresponding to the printing machine, the chip mounter, and the reflow soldering in real time, and store the initial operation data in a preset database; the initial operation data includes corresponding core parameters and production information;

[0034] An operation data processing module, configured to perform preprocessing operations on each of the initial operation data in the preset database to obtain target operation data, and use a number of machine learning models to process each of the target operation data to obtain an initial parameter prediction result; the initial parameter prediction result includes a first-pass rate prediction result, a theoretical time-consuming prediction result, and a printing speed prediction result;

[0035] A parameter prediction result determination module, configured to call a preset reinforcement learning model and use the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to process the initial parameter prediction result to obtain a corresponding target parameter prediction result, so as to configure the device using the target parameter prediction result.

[0036] In a third aspect, the present application provides an electronic device, including:

[0037] A memory, configured to store a computer program;

[0038] A processor, configured to execute the computer program to implement the foregoing device parameter prediction method based on machine learning.

[0039] In a fourth aspect, the present application provides a computer-readable storage medium, configured to store a computer program, wherein the computer program, when executed by a processor, implements the foregoing device parameter prediction method based on machine learning.

[0040] As can be seen from the above, before predicting the device parameters based on machine learning in this application, it is necessary to use preset sensors and data acquisition devices to collect the initial operation data corresponding to the printer, mounter, and reflow soldering machine in real time, and store the initial operation data in a preset database; the initial operation data includes the corresponding core parameters and production information; perform preprocessing operations on each of the initial operation data in the preset database to obtain target operation data, and use several machine learning models to process each of the target operation data to obtain initial parameter prediction results; the initial parameter prediction results include the first-pass yield prediction result, theoretical time consumption prediction result, and printing speed prediction result; call a preset reinforcement learning model and use the deep deterministic policy gradient algorithm and proximal policy optimization algorithm to process the initial parameter prediction results to obtain corresponding target parameter prediction results, so as to configure the device using the target parameter prediction results.

[0041] As can be seen, this application first needs to collect the initial operation data corresponding to the printer, mounter, and reflow soldering machine in real time by using preset sensors and data acquisition devices, and store the initial operation data in a preset database; the initial operation data includes the corresponding core parameters and production information; subsequently, perform preprocessing operations on each of the initial operation data in the preset database to obtain target operation data, and use several machine learning models to process each of the target operation data to obtain initial parameter prediction results; the initial parameter prediction results include the first-pass yield prediction result, theoretical time consumption prediction result, and printing speed prediction result; finally, call a preset reinforcement learning model and use the deep deterministic policy gradient algorithm and proximal policy optimization algorithm to process the initial parameter prediction results to obtain corresponding target parameter prediction results, so as to configure the device using the target parameter prediction results. In this way, the quality and speed of obtaining process parameters are improved, and thus the production efficiency of the production process is enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0043] Figure 1 It is a flowchart of a method for predicting device parameters based on machine learning disclosed in this application;

[0044] Figure 2 It is a schematic structural diagram of a device for predicting device parameters based on machine learning disclosed in this application;

[0045] Figure 3 A structural diagram of an electronic device disclosed in this application. Specific implementation manners

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0047] With the intensification of market competition, manufacturing enterprises are facing various challenges, including improving production efficiency, reducing production costs, and enhancing product quality. In the traditional manufacturing mode, the setting of process parameters of equipment often lacks a scientific basis, resulting in unstable product quality. This situation restricts the sustainable development of enterprises and also hinders the process of intelligent manufacturing to a certain extent. For this reason, more and more enterprises have begun to try to apply machine learning algorithms to the analysis and prediction of production data. However, the current application of machine learning algorithms is still in its infancy, and there is a lack of systematic analysis of the relationships between different devices and different process parameters, making the construction and application of machine learning models face many challenges. For this reason, this application provides a method for predicting equipment parameters based on machine learning, which can improve the quality and speed of obtaining process parameters, and thus enhance the efficiency of the production process.

[0048] See Figure 1 As shown, the embodiments of the present invention disclose a method for predicting equipment parameters based on machine learning, including:

[0049] Step S11: Use a preset sensor and a data acquisition device to collect the initial operation data corresponding to a printer, a mounter, and a reflow oven in real time, and store the initial operation data in a preset database; the initial operation data includes the corresponding core parameters and production information.

[0050] In this embodiment, the data acquisition center is the basis of this application embodiment. It mainly obtains the corresponding operation status from each running device in the production process through sensors and data acquisition devices. Specifically, the use of a preset sensor and a data acquisition device to collect the initial operation data corresponding to a printer, a mounter, and a reflow oven in real time may include: using a preset sensor to collect the printing speed, demolding time, printing pressure, and printing temperature corresponding to the printer in real time; using a data acquisition device to collect the mounter speed, mounting accuracy, mounter production volume, mounter operation time, and mounter failure information corresponding to the mounter in real time; using a preset sensor to collect the welding temperature curve, welding time, and welding furnace atmosphere corresponding to the reflow oven in real time.

[0051] Step S12: Perform preprocessing operations on each of the initial operation data in the preset database to obtain target operation data, and use a number of machine learning models to process each of the target operation data to obtain initial parameter prediction results; the initial parameter prediction results include a first-pass yield prediction result, a theoretical time consumption prediction result, and a printing speed prediction result.

[0052] In this embodiment, preprocessing the data is a key step to ensure the accuracy of the model output results. In a specific implementation manner, it includes operations such as removing missing values, outliers, and duplicate data from the data to perform data cleaning on the collected initial operation data, thereby ensuring the quality of the data set. In addition, the embodiments of the present application need to perform standardization processing on the initial operation data with different dimensions so that they are within the same range for subsequent model training. Finally, by combining and transforming each device parameter, useful features are extracted to improve the parameter prediction ability of the model. Specifically, the step of performing preprocessing operations on each of the initial operation data in the preset database to obtain target operation data may include: performing operations such as removing missing values, outliers, and duplicate data on each of the initial operation data in the preset database to obtain cleaned operation data; performing standardization processing on each of the cleaned operation data using a preset standardization operation to obtain standardized data; performing combination and transformation operations on each of the standardized data using a preset row combination and transformation operation to obtain data to be extracted, and performing feature extraction operations on the data to be extracted using a preset feature extraction method to obtain target operation data.

[0053] Further, since the embodiments of the present application need to predict device parameters corresponding to different devices, therefore, the embodiments of the present application select different machine learning models to predict the corresponding prediction targets: First, predict the first-pass yield of the device. In a specific implementation manner, the core parameters and production data of the device are used as input features of the decision tree model, and the first-pass yield is used as the target variable. Then, the cleaned data set is divided into a training set and a test set to train the decision tree model through the decision tree algorithm, and hyperparameters are optimized during the process of training the decision tree model, such as the depth of the tree and the minimum sample split number. It is worth mentioning that after each training of the decision tree model, the embodiments of the present application need to evaluate the performance of the trained model through cross-validation and preset metrics, such as the root mean square error RMSE, to ensure that the model can predict the first-pass yield in real time according to the current device state.

[0054] Specifically, the process of using several machine learning models to process each piece of the target operation data to obtain the initial parameter prediction result may include: dividing each piece of the target operation data according to a preset division rule to obtain a training set and a test set, and then training an initial decision tree regression model based on the training set and the test set to obtain an evaluation-to-be decision tree regression model; using a preset cross-validation method and a preset model metric to determine whether the evaluation-to-be decision tree regression model meets the preset performance condition. If the evaluation-to-be decision tree regression model meets the preset performance condition, setting the evaluation-to-be decision tree regression model as the target decision tree regression model, and using the target decision tree regression model to process each piece of the target operation data to obtain the corresponding throughput rate prediction result.

[0055] Secondly, predict the theoretical time consumption of the equipment. In a specific implementation, the core parameters corresponding to the printing process, the chip mounting process, and the soldering process are used as input features, and the theoretical time consumption is used as the target variable. Subsequently, the least squares method is used to train the linear regression model to obtain the time consumption prediction formula corresponding to each process during the training of the linear regression model. Then, the time consumption prediction formula corresponding to each process is used to process each equipment parameter to obtain the theoretical time consumption corresponding to each process, and compare it with the actual time consumption to help identify the production bottleneck in the production process.

[0056] Specifically, the process of using several machine learning models to process each piece of the target operation data to obtain the initial parameter prediction result may include: using the least squares method and based on the core parameters corresponding to the printing process, the chip mounting process, and the soldering process respectively to train each initial time consumption prediction formula in the initial linear regression model to obtain the target time consumption prediction formula corresponding to the printing process, the chip mounting process, and the soldering process respectively; using each target time consumption prediction formula to perform a theoretical time consumption calculation operation on the corresponding core parameter to obtain the corresponding theoretical time consumption prediction result.

[0057] Finally, predict the process parameters of the equipment. In a specific implementation, the front and rear squeegee pressures are used as input features, and the printing speed parameter is used as the target variable. Subsequently, an appropriate K value is set based on user requirements to train the initial printing speed prediction model through the K-nearest neighbor algorithm based on the determined K value and historical data, so as to predict the printing speed of the printing press using the obtained target printing speed prediction model. Specifically, the use of several machine learning models to process each of the target operation data to obtain an initial parameter prediction result may include: setting a target K value based on user requirements to train the initial printing speed prediction model using the K-nearest neighbor regression algorithm based on the target K value and the front and rear squeegee pressures corresponding to the printing press to obtain a target printing speed prediction model; using the target printing speed prediction model to predict the printing speed of the printing press to obtain a printing speed prediction result.

[0058] Step S13: Invoke a preset reinforcement learning model and use the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to process the initial parameter prediction result to obtain a corresponding target parameter prediction result, so as to configure the equipment using the target parameter prediction result.

[0059] In this embodiment, after obtaining the initial parameter prediction result, the embodiment of the present application needs to introduce a reinforcement learning algorithm and perform adaptive process optimization based on the prediction result to improve the production efficiency and stability in the production process. First, the embodiment of the present application needs to reinforce the learning agent through the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to learn the optimal process parameter adjustment strategy. Subsequently, the corresponding action to be executed is determined based on the state composed of the first-pass yield prediction result, the solder joint quality prediction result, and the production beat prediction result. The actions to be executed include process parameters such as adjusting the printing speed, placement accuracy, welding temperature, and equipment operation rate. Finally, by comprehensively considering factors such as product qualification rate, production time, energy consumption, and equipment health, a multi-objective optimization reward function is established to guide the reinforcement learning model to converge to the optimal production strategy.

[0060] Further, the embodiments of the present application adopt a multi-objective optimization strategy to optimize the initial parameter prediction results. Among them, the optimization objectives include but are not limited to quality, efficiency, energy consumption, and equipment life. In a specific embodiment, the embodiments of the present application adopt a weighted comprehensive scoring strategy to optimize the initial parameter prediction results, that is, dynamically adjust the weights according to business requirements to balance production efficiency and equipment wear. In another specific embodiment, the embodiments of the present application adopt the Pareto Optimal Solution Method (Pareto Optimal Solution Method, that is, an optimal solution algorithm) to optimize the initial parameter prediction results to find the optimal process parameter combination that satisfies the trade-offs in multiple aspects such as quality-cost-time.

[0061] Specifically, the process of calling a preset reinforcement learning model and using the Deep Deterministic Policy Gradient algorithm and the Proximal Policy Optimization algorithm to process the initial parameter prediction results to obtain corresponding target parameter prediction results may include: calling a preset reinforcement learning model and using the Deep Deterministic Policy Gradient algorithm and the Proximal Policy Optimization algorithm to process the state to be processed to obtain corresponding actions to be executed; the state to be processed includes the first-pass yield prediction result, the theoretical time consumption prediction result, and the printing speed prediction result; adjusting the process parameters corresponding to the printer, the mounter, and the reflow oven based on each of the actions to be executed to obtain a reward value corresponding to the action to be executed; the process parameters include printing speed, placement accuracy, soldering temperature, and equipment operation rate; using a preset weighted comprehensive scoring rule and the Pareto Optimal Solution Method and based on the reward value to adjust the parameters of the preset reinforcement learning model, so as to use the adjusted preset reinforcement learning model to process the initial parameter prediction results to obtain corresponding target parameter prediction results.

[0062] Thus, the present application first needs to collect the initial operation data corresponding to the printer, the mounter, and the reflow oven in real time by using a preset sensor and a data acquisition device, and store the initial operation data in a preset database; the initial operation data includes the corresponding core parameters and production information; subsequently, perform a preprocessing operation on each piece of initial operation data in the preset database to obtain target operation data, and use a number of machine learning models to process each piece of target operation data to obtain initial parameter prediction results; the initial parameter prediction results include the first-pass yield prediction result, the theoretical time consumption prediction result, and the printing speed prediction result; finally, call a preset reinforcement learning model and use the Deep Deterministic Policy Gradient algorithm and the Proximal Policy Optimization algorithm to process the initial parameter prediction results to obtain corresponding target parameter prediction results, so as to configure the equipment by using the target parameter prediction results. In this way, the quality and speed of obtaining process parameters are improved, and thus the production efficiency of the production process is enhanced.

[0063] Correspondingly, referring to Figure 2 as shown, the present application also provides a device parameter prediction device based on machine learning, including:

[0064] An operation data acquisition module 11, configured to use a preset sensor and a data acquisition device to collect initial operation data corresponding to a printing machine, a chip mounter, and a reflow soldering in real time, and store the initial operation data in a preset database; the initial operation data includes corresponding core parameters and production information;

[0065] An operation data processing module 12, configured to perform a preprocessing operation on each of the initial operation data in the preset database to obtain target operation data, and use a plurality of machine learning models to process each of the target operation data to obtain an initial parameter prediction result; the initial parameter prediction result includes a first-pass rate prediction result, a theoretical time consumption prediction result, and a printing speed prediction result;

[0066] A parameter prediction result determination module 13, configured to call a preset reinforcement learning model and use a deep deterministic policy gradient algorithm and a proximal policy optimization algorithm to process the initial parameter prediction result to obtain a corresponding target parameter prediction result, so as to configure the device by using the target parameter prediction result.

[0067] As can be seen from the above, before performing device parameter prediction based on machine learning in the embodiments of the present application, it is first necessary to use a preset sensor and a data acquisition device to collect initial operation data corresponding to a printing machine, a chip mounter, and a reflow soldering in real time, and store the initial operation data in a preset database; the initial operation data includes corresponding core parameters and production information; subsequently, perform a preprocessing operation on each of the initial operation data in the preset database to obtain target operation data, and use a plurality of machine learning models to process each of the target operation data to obtain an initial parameter prediction result; the initial parameter prediction result includes a first-pass rate prediction result, a theoretical time consumption prediction result, and a printing speed prediction result; finally, call a preset reinforcement learning model and use a deep deterministic policy gradient algorithm and a proximal policy optimization algorithm to process the initial parameter prediction result to obtain a corresponding target parameter prediction result, so as to configure the device by using the target parameter prediction result. In this way, the quality and speed of obtaining process parameters are improved, and thus the production efficiency of the production process is enhanced.

[0068] In some specific embodiments, the operation data acquisition module 11 may specifically include:

[0069] A first operation data acquisition subunit, configured to use a preset sensor to collect the printing speed, demolding time, printing pressure, and printing temperature corresponding to the printing machine in real time;

[0070] The second operation data acquisition subunit is used to use a data acquisition device to collect in real time the chip mounting speed, mounting accuracy, chip production volume, operation time of the chip mounter, and chip mounter failure information corresponding to the chip mounter;

[0071] The third operation data acquisition subunit is used to use a preset sensor to collect in real time the welding temperature curve, welding time, and atmosphere in the welding furnace corresponding to the reflow soldering.

[0072] In some specific embodiments, the operation data processing module 12 may specifically include:

[0073] The operation data cleaning unit is used to perform operations of removing missing values, outliers, and duplicate data on each of the initial operation data in the preset database to obtain the cleaned operation data;

[0074] The operation data standardization processing unit is used to perform standardization processing on each of the cleaned operation data by using a preset standardization operation to obtain standardized data;

[0075] The feature extraction unit is used to perform combination and transformation operations on each of the standardized data by using a preset row combination and transformation operation to obtain the data to be extracted, and perform feature extraction operations on the data to be extracted by using a preset feature extraction method to obtain the target operation data.

[0076] In some specific embodiments, the operation data processing module 12 may specifically include:

[0077] The first model training unit is used to divide each of the target operation data according to a preset division rule to obtain a training set and a test set, and then train an initial decision tree regression model based on the training set and the test set to obtain a decision tree regression model to be evaluated;

[0078] The first pass rate prediction result determination unit is used to use a preset cross-validation method and a preset model metric to determine whether the decision tree regression model to be evaluated meets the preset performance conditions. If the decision tree regression model to be evaluated meets the preset performance conditions, the decision tree regression model to be evaluated is set as the target decision tree regression model, and the target decision tree regression model is used to process each of the target operation data to obtain the corresponding first pass rate prediction result.

[0079] In some specific embodiments, the operation data processing module 12 may specifically include:

[0080] A second model training unit, configured to use the least squares method and train each initial time-consuming prediction formula in the initial linear regression model based on the core parameters corresponding to the printing process, the chip mounting process, and the soldering process respectively, to obtain target time-consuming prediction formulas corresponding to the printing process, the chip mounting process, and the soldering process respectively;

[0081] A theoretical time-consuming prediction result determination unit, configured to use each of the target time-consuming prediction formulas to perform a theoretical time-consuming calculation operation on the corresponding core parameters, to obtain corresponding theoretical time-consuming prediction results.

[0082] In some specific embodiments, the operation data processing module 12 may specifically include:

[0083] A third model training unit, configured to set a target K value based on user requirements, so as to use the K-nearest neighbor regression algorithm and train the initial printing speed prediction model based on the target K value and the front and rear squeegee pressures corresponding to the printing machine, to obtain a target printing speed prediction model;

[0084] A printing speed prediction result determination unit, configured to use the target printing speed prediction model to predict the printing speed of the printing machine, to obtain a printing speed prediction result.

[0085] In some specific embodiments, the parameter prediction result determination module 13 may specifically include:

[0086] A to-be-executed action determination unit, configured to call a preset reinforcement learning model and use the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to process the to-be-processed state, to obtain a corresponding to-be-executed action; the to-be-processed state includes the straight-through rate prediction result, the theoretical time-consuming prediction result, and the printing speed prediction result;

[0087] A reward value determination unit, configured to adjust the process parameters corresponding to the printing machine, the chip mounter, and the reflow soldering based on each of the to-be-executed actions, to obtain a reward value corresponding to the to-be-executed action; the process parameters include printing speed, chip mounting accuracy, soldering temperature, and equipment operation rate;

[0088] A model parameter adjustment unit, configured to use a preset weighted comprehensive scoring rule and the Pareto optimal solution method and adjust the parameters of the preset reinforcement learning model based on the reward value, so as to use the adjusted preset reinforcement learning model to process the initial parameter prediction result, to obtain a corresponding target parameter prediction result.

[0089] Furthermore, an embodiment of the present application also discloses an electronic device, Figure 3It is a structural diagram of an electronic device 20 shown according to an exemplary embodiment. The content in the figure should not be considered as any limitation on the scope of use of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the device parameter prediction method based on machine learning disclosed in any of the foregoing embodiments. Additionally, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0090] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and no specific limitation is imposed on it here; the input / output interface 25 is used to obtain external input data or output data to the outside, and its specific interface type can be selected according to specific application needs, and no specific limitation is made here.

[0091] In addition, as a carrier for resource storage, the memory 22 can be a read-only memory, a random access memory, a magnetic disk, or an optical disc, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0092] Among them, the operating system 221 is used to manage and control each hardware device and the computer program 222 on the electronic device 20, and it can be Windows Server, Netware, Unix, Linux, etc. The computer program 222, in addition to including a computer program that can be used to complete the device parameter prediction method based on machine learning executed by the electronic device 20 disclosed in any of the foregoing embodiments, may further include a computer program that can be used to complete other specific tasks.

[0093] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the device parameter prediction method based on machine learning disclosed above. For the specific steps of this method, reference can be made to the corresponding content disclosed in the foregoing embodiments, and details will not be repeated here.

[0094] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0095] Those skilled in the art may further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0096] The steps of the methods or algorithms described in combination with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of both. The software modules can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the technical field.

[0097] Finally, it should also be noted that in this document, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising a..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0098] The technical solutions provided in this application have been introduced in detail above. Specific examples have been used herein to elaborate on the principles and implementation manners of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A device parameter prediction method based on machine learning, characterized in that: include: Using preset sensors and data acquisition equipment to collect initial operation data corresponding to the printer, the placement machine and the reflow soldering machine in real time, and storing the initial operation data in a preset database; The initial operation data includes respectively corresponding core parameters and output information; Preprocessing each of the initial operation data in the preset database to obtain target operation data, and using a number of machine learning models to process each of the target operation data to obtain an initial parameter prediction result; The initial parameter prediction results include a straight-through rate prediction result, a theoretical time-consuming prediction result, and a printing speed prediction result; The preset reinforcement learning model is called and the initial parameter prediction results are processed using a deep deterministic policy gradient algorithm and a proximal policy optimization algorithm to obtain corresponding target parameter prediction results, so as to configure the device using the target parameter prediction results.

2. The device parameter prediction method based on machine learning according to claim 1 is characterized in that: The method of using the preset sensors and data acquisition equipment to collect the initial operation data corresponding to the printer, the placement machine and the reflow soldering machine in real time includes: Using preset sensors to collect the printing speed, demoulding time, printing pressure and printing temperature corresponding to the printing machine in real time; Using data acquisition equipment to collect the chip placement speed, placement accuracy, chip production volume, chip placement machine operation time and chip placement machine fault information corresponding to the chip placement machine in real time; The preset sensor is used to collect the welding temperature curve, welding time and atmosphere in the welding furnace corresponding to the reflow soldering in real time.

3. The device parameter prediction method based on machine learning according to claim 1 is characterized in that: The preprocessing operation is performed on each of the initial operation data in the preset database to obtain target operation data, including: Removing missing values, outliers and duplicate data from each of the initial operating data in the preset database to obtain cleaned operating data; Performing standardization processing on each of the post-cleaning operation data using a preset standardization operation to obtain standardized data; The standardized data are combined and transformed using preset row combination and transformation operations to obtain data to be extracted, and feature extraction operations are performed on the data to be extracted using a preset feature extraction method to obtain target operation data.

4. The device parameter prediction method based on machine learning according to claim 1 is characterized in that: The method of using a plurality of machine learning models to process the target operation data to obtain initial parameter prediction results includes: Dividing each of the target operation data according to a preset division rule to obtain a training set and a test set, and then training an initial decision tree regression model based on the training set and the test set to obtain a decision tree regression model to be evaluated; A preset cross-validation method and preset model indicators are used to determine whether the decision tree regression model to be evaluated meets the preset performance conditions. If the decision tree regression model to be evaluated meets the preset performance conditions, the decision tree regression model to be evaluated is set as the target decision tree regression model, and the target decision tree regression model is used to process each of the target operation data to obtain the corresponding pass rate prediction result.

5. The device parameter prediction method based on machine learning according to claim 1 is characterized in that: The method of using a plurality of machine learning models to process the target operation data to obtain initial parameter prediction results includes: The initial time-consuming prediction formulas in the initial linear regression model are trained by using the least squares method based on the core parameters corresponding to the printing process, the patch process and the welding process, respectively, to obtain the target time-consuming prediction formulas corresponding to the printing process, the patch process and the welding process, respectively; The theoretical time consumption calculation operation is performed on the corresponding core parameters using each of the target time consumption prediction formulas to obtain the corresponding theoretical time consumption prediction results.

6. The device parameter prediction method based on machine learning according to claim 1 is characterized in that: The method of using a plurality of machine learning models to process the target operation data to obtain initial parameter prediction results includes: A target K value is set based on user needs, so as to train an initial printing speed prediction model based on the target K value and the front and rear scraper pressures corresponding to the printing press using a K nearest neighbor regression algorithm to obtain a target printing speed prediction model; The target printing speed prediction model is used to predict the printing speed of the printing press to obtain a printing speed prediction result.

7. The device parameter prediction method based on machine learning according to any one of claims 1 to 6, characterized in that: The calling of the preset reinforcement learning model and using the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to process the initial parameter prediction results to obtain the corresponding target parameter prediction results include: Calling a preset reinforcement learning model and using a deep deterministic policy gradient algorithm and a proximal policy optimization algorithm to process the state to be processed, and obtaining a corresponding action to be executed; the state to be processed includes the first pass rate prediction result, the theoretical time consumption prediction result, and the printing speed prediction result; Adjust the process parameters corresponding to the printer, the placement machine and the reflow soldering based on each of the actions to be performed, and obtain a reward value corresponding to the actions to be performed; the process parameters include printing speed, placement accuracy, welding temperature and equipment operation rate; The preset weighted comprehensive scoring rule and the Pareto optimal solution method are used to adjust the parameters of the preset reinforcement learning model based on the reward value, so as to process the initial parameter prediction results using the adjusted preset reinforcement learning model to obtain corresponding target parameter prediction results.

8. A device parameter prediction device based on machine learning, characterized in that: include: An operation data acquisition module is used to use preset sensors and data acquisition equipment to collect the initial operation data corresponding to the printer, the placement machine and the reflow soldering machine in real time, and store the initial operation data in a preset database; The initial operation data includes respectively corresponding core parameters and output information; An operation data processing module is used to perform a preprocessing operation on each of the initial operation data in the preset database to obtain target operation data, and use a number of machine learning models to process each of the target operation data to obtain an initial parameter prediction result; the initial parameter prediction result includes a straight pass rate prediction result, a theoretical time consumption prediction result, and a printing speed prediction result; The parameter prediction result determination module is used to call the preset reinforcement learning model and use the deep deterministic policy gradient algorithm and the proximal policy optimization algorithm to process the initial parameter prediction results to obtain the corresponding target parameter prediction results, so as to configure the device using the target parameter prediction results.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the device parameter prediction method based on machine learning as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the steps of the device parameter prediction method based on machine learning as described in any one of claims 1 to 7 are implemented.

Images