Metal processing prediction and processing method and system based on deep neural network

Abstract

The invention provides a metal processing prediction and processing method and system based on a deep neural network, and the method comprises the steps: obtaining all machining parameters and machining results in each metal processing process, and enabling the machining parameters, the machining results, corresponding raw material parameters and corresponding quality scoring parameters to serve as metal processing data; constructing a DNN model, inputting the metal processing data into the constructed DNN model for training, and obtaining and storing a trained metal processing prediction model; loading a metal processing prediction model, and inputting the current raw material parameters, the current equipment state parameters and the current quality scoring parameters of the current product into the metal processing prediction model to obtain to-be-optimized equipment setting parameters of the next product output by the metal processing prediction model; and before the production of the next product is started, the equipment setting parameters are adjusted according to the to-be-optimized equipment setting parameters. The machining parameters can be adjusted, the machining technology is optimized, and precise technology control is achieved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and industrial automation technology, and in particular to a metal processing prediction and processing method and system based on a deep neural network. Background Art

[0002] With the advancement of Industry 4.0, intelligent manufacturing has become a trend in industrial development. For example, artificial intelligence models are applied to the field of metal processing, such as the application number CN202310008813.4, a metal processing surface defect detection system and detection method based on shape features, which combines traditional image positioning and segmentation algorithms with convolutional neural networks to obtain metal processing surface defect classification maps, that is, the existing artificial intelligence model is to automatically recognize the results of metal processing images, while the metal processing process still relies on experience, and it is difficult to achieve precise process control and quality assurance. Summary of the invention

[0003] In order to solve the above-mentioned problems in the prior art, the present invention provides a metal processing prediction and processing method and system based on a deep neural network, which can achieve precise process control.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is:

[0005] In a first aspect, the present invention provides a metal processing prediction and processing method based on a deep neural network, comprising the following steps:

[0006] Obtaining various processing parameters and processing results in each metal processing process, and using the processing parameters, the processing results, the corresponding raw material parameters and the corresponding quality scoring parameters as metal processing data, wherein the processing parameters include equipment state parameters and equipment setting parameters, wherein the equipment state parameters refer to the state parameters of the metal processing equipment during the metal processing process, and the equipment setting parameters refer to the production parameters that need to be set in the metal processing equipment;

[0007] Constructing a DNN model, inputting the metal processing data into the constructed DNN model for training, obtaining and saving the trained metal processing prediction model, using the raw material parameters, the equipment status parameters and the quality scoring parameters as input parameters of the DNN model, and selecting new equipment setting parameters as output parameters of the DNN model according to the processing results and the quality scoring parameters;

[0008] Loading the metal processing prediction model, inputting the current raw material parameters, current equipment status parameters and current quality score parameters of the current product into the metal processing prediction model, and obtaining the equipment setting parameters to be optimized for the next product output by the metal processing prediction model;

[0009] Before the production of the next product starts, the equipment setting parameters are adjusted according to the equipment setting parameters to be optimized.

[0010] The beneficial effects of the present invention are: obtaining various processing parameters and processing results in each metal processing process, and using them as training data to train the DNN model to obtain a metal processing prediction model, thereby using the metal processing prediction model to perform real-time prediction of new processing data, and adjusting the processing parameters according to the prediction results, optimizing the processing technology, and achieving precise process control.

[0011] Optionally, the selecting new equipment setting parameters as output parameters of the DNN model according to the processing results and the quality score parameters comprises the following steps:

[0012] For a first input parameter, obtain a processing data set SQ that belongs to the same type of product and the same production line as the first input parameter and whose processing results reach the quality score parameter, and select a set of equipment setting parameters that best matches the first equipment setting parameters in the first metal processing data where the first input parameter is located as new equipment setting parameters from the processing data set SQ to obtain the output parameters of the first input parameter in the DNN model.

[0013] It can be known from the above description that new device setting parameters are automatically set by self-matching the device setting parameters.

[0014] Optionally, the matching relationship between the two groups of device setting parameters is obtained by accumulating the product of the difference between each setting parameter in the two groups of device setting parameters and the corresponding weight value.

[0015] Optionally, the method further comprises the following steps:

[0016] If the amount of data in the processing data set SQ is lower than the lower limit threshold, the metal processing data of the same type of products and different production lines, and the metal processing data of the same production line but different types of products are added in turn, among which the conversion ratios of the metal processing data of the same type of products and the same production line, the metal processing data of the same type of products and different production lines, and the metal processing data of the same production line but different types of products are reduced in turn when matching equipment setting parameters.

[0017] Optionally, the next product is the next product from the same batch as the current product or the product from the next batch of the current product.

[0018] Optionally, before constructing the DNN model, the steps of:

[0019] Performing data preprocessing on the metal processing data, wherein the data preprocessing includes standardization processing, missing data filling and abnormal data processing;

[0020] Before inputting the current raw material parameters, current equipment status parameters and current quality score parameters of the current product as input parameters into the metal processing prediction model, the following steps are also included:

[0021] The current raw material parameters, current equipment status parameters and current quality score parameters of the current product are preprocessed.

[0022] Optionally, the data preprocessing is implemented using Python's Pandas library.

[0023] Optionally, after constructing the DNN model, the following steps are further included:

[0024] Define the hyperparameter search space and use Bayesian optimization methods to tune the hyperparameters of the DNN model.

[0025] Optionally, the equipment status parameters include equipment core component temperature, main power voltage, core roller vibration amplitude and air supply pressure;

[0026] The raw material parameters include the raw material appearance size, the alloy ratio inside the raw material, the raw material initial temperature, and the material surface oil film thickness;

[0027] The quality scoring parameters include product requirement parameters in the customer order, product surface flatness, average thickness and number of defects;

[0028] The equipment setting parameters include the target temperature value for preheating the incoming materials, the set pressure under the equipment core and the advancing distance value for the incoming materials.

[0029] Among them, the technical effects corresponding to the metal processing prediction and processing system based on deep neural network provided by the second aspect refer to the relevant description of the metal processing prediction and processing method based on deep neural network provided by the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic flow chart of a metal processing prediction and processing method based on a deep neural network according to an embodiment of the present invention;

[0031] Figure 2 A schematic diagram of a process of model training according to an embodiment of the present invention;

[0032] Figure 3 A schematic diagram of a flow chart of model prediction involved in an embodiment of the present invention;

[0033] Figure 4 It is a structural schematic diagram of a metal processing prediction and processing method based on a deep neural network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to better understand the above technical solution, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0035] Embodiment 1

[0036] Please refer to Figure 1 , a metal processing prediction and processing method based on deep neural network includes the following steps:

[0037] S1. Obtain various processing parameters and processing results in each metal processing process, and use the processing parameters, processing results, corresponding raw material parameters and corresponding quality score parameters as metal processing data. The processing parameters include equipment status parameters and equipment setting parameters. The equipment status parameters refer to the status parameters of the metal processing equipment during the metal processing process, and the equipment setting parameters refer to the production parameters that need to be set in the metal processing equipment.

[0038] In this embodiment, the input parameters are the external environment and uncontrollable factors and fixed parameters. Among them, the equipment status parameters include the temperature of the core components of the equipment, the main power voltage, the vibration amplitude of the core roller and the air supply pressure. The raw material parameters include the appearance size of the raw material, the alloy ratio inside the raw material, the initial temperature of the raw material, and the oil film thickness on the surface of the material. The quality scoring parameters include the product requirement parameters in the customer order, the surface flatness of the product, the average thickness and the number of defects. The equipment setting parameters include the target temperature value for preheating the incoming materials, the set pressure under the core of the equipment and the advancing distance value of the incoming materials. Among them, the above parameters will vary according to different metal processing processes.

[0039] In this embodiment, before executing step S2, the following steps are also included:

[0040] Preprocess the metal processing data, including standardization, missing data filling and abnormal data processing;

[0041] Among them, the standardization process is to eliminate the scale difference of the data, and the missing data filling and abnormal data processing are to ensure the data quality. In this embodiment, the data preprocessing is implemented using the Python Pandas library, and the specific implementation is as follows:

[0042] import pandas as pdfromsklearn.preprocessing import StandardScaler

[0043] #Import of data collection results

[0044] data=pd.read_csv(' / path / to / data.csv')

[0045] #Data preprocessing

[0046] scaler = StandardScaler()

[0047] data_scaled=scaler.fit_transform(data)

[0048] S2. Construct a DNN model, input metal processing data into the constructed DNN model for training, obtain and save the trained metal processing prediction model, use raw material parameters, equipment status parameters and quality score parameters as input parameters of the DNN model, and select new equipment setting parameters as output parameters of the DNN model based on the processing results and quality score parameters.

[0049] Among them, the DNN model includes input layer, hidden layer and output layer, and the specific implementation is as follows:

[0050] a) Implementation of data initialization and reverse transfer of the fully connected layer

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057] In this embodiment, after constructing the DNN model, the following steps are also included:

[0058] Define the hyperparameter search space and use Bayesian optimization methods to tune the hyperparameters of the DNN model.

[0059] Therefore, the Bayesian optimization method is used to optimize the hyperparameters to improve the prediction accuracy of the model. The specific implementation process of Bayesian optimization is as follows:

[0060] #Define hyperparameter space

[0061] hyper_dim_space={

[0062] 'hidden_dim1': Integer(32, 128),

[0063] 'hidden_dim2': Integer(32, 128),

[0064] 'learning_rate': Real(0.0001, 0.01, prior='log-uniform'),

[0065] 'batch_size': Integer(10, 100),

[0066] 'epochs': Integer(50, 200)

[0067] }

[0068] # Automatically select and apply different search methods

[0069] def auto_select_search_method(hyper_dim_space, X_train, y_train, scoring, n_iter = 20):

[0070] n_samples, n_features = X_train.shape

[0071] if n_samples * n_features < 1000:

[0072] print("Using Grid Search")

[0073] search = GridSearchCV(estimator = DNNRegressor(), param_grid = hyper_dim_space, cv = 3, scoring = scoring)

[0074] elif n_samples * n_features < 5000:

[0075] print("Using Randomized Search")

[0076] search=RandomizedSearchCV(estimator=DNNRegressor(),param_distributions=hyper_dim_space,n_iter=n_iter,cv=3,scoring=scoring,random_state=42)

[0077] elifn_samples*n_features<10000:

[0078]

[0079] In this embodiment, selecting new equipment setting parameters as output parameters of the DNN model according to the processing results and the quality score parameters includes the following steps:

[0080] For the first input parameter, obtain a processing data set SQ that belongs to the same type of product and the same production line as the first input parameter and whose processing results reach the quality score parameter, and select a set of equipment setting parameters that best matches the first equipment setting parameters in the first metal processing data where the first input parameter is located as new equipment setting parameters from the processing data set SQ to obtain the output parameters of the first input parameter in the DNN model.

[0081] Among them, the matching relationship between the two sets of equipment setting parameters is obtained by accumulating the product of the difference between each setting parameter in the two sets of equipment setting parameters and the corresponding weight value. The setting of the above weight value is based on the degree of influence of the setting parameter on the final result and the difficulty of adjusting the setting parameter itself. For example, the adjustment response speed of some setting parameters is slow and cannot be reflected in the production process in time, and the adjustment operation of some setting parameters is more troublesome, etc.

[0082] Among them, according to the relationship between the processing results and the quality scoring parameters, it is possible to know whether this group of processing parameters meets the requirements, so as to select the setting parameters that are easy to adjust and have obvious adjustment effects as the output parameters, so that the setting parameters produced by the model during training are also easier to adjust and have obvious adjustment effects.

[0083] In this embodiment, the following steps are also included:

[0084] If the amount of data in the processing data set SQ is lower than the lower limit threshold, the metal processing data of the same type of products but different production lines and the metal processing data of the same production line but different types of products are added in turn, among which the conversion ratios of the metal processing data of the same type of products and the same production line, the metal processing data of the same type of products but different production lines, and the metal processing data of the same production line but different types of products are reduced in turn when matching equipment setting parameters.

[0085] Among them, as the conversion ratio decreases, the corresponding data requires a higher matching value to be selected during matching, so that when different production lines or different types of products are introduced, the production lines and product types are also taken into consideration in the data matching process.

[0086] Reference Figure 2 It can be seen that in the present embodiment, during the training process, the model is verified to determine whether it has been trained well. If so, it is constructed and saved as a metal processing prediction model. Therefore, the specific implementation process of model training is as follows:

[0087] #Retrain the model using the best hyperparameters

[0088] best_model=DNNRegressor(**hyper_params).fit(X_train_scaled,y_train_scaled)

[0089] #Save the model and related data

[0090] torch.save(best_model.model.state_dict(),'best_model.pth')

[0091] with open('scalers.pkl','wb')as f:

[0092] pickle.dump({'scaler_X':scaler_X,'scaler_y':scaler_y,'hyper_params':hyper_params},f)

[0093] S3, loading the metal processing prediction model, inputting the current raw material parameters, current equipment status parameters and current quality score parameters of the current product into the metal processing prediction model, and obtaining the equipment setting parameters to be optimized for the next product output by the metal processing prediction model;

[0094] In this embodiment, standard sensors and data acquisition devices are used to collect various parameters during the processing in real time.

[0095] Among them, the next product is the next product from the same batch of the current product, the piston, the product from the next batch of the current product.

[0096] Before inputting the current raw material parameters, current equipment status parameters and current quality score parameters of the current product as input parameters into the metal processing prediction model, the following steps are also included:

[0097] The current raw material parameters, current equipment status parameters and current quality score parameters of the current product are preprocessed.

[0098] In this embodiment, the specific implementation process of the model prediction is as follows:

[0099]

[0100]

[0101] Reference Figure 3 It can be seen that after the model prediction, its prediction results must be de-normalized to obtain data that the user can actually understand.

[0102] S4. Before the production of the next product begins, the equipment setting parameters are adjusted according to the equipment setting parameters to be optimized.

[0103] In the metal processing process, real-time production data can reflect the effect of the finished product. Therefore, the real-time production data of the current product is used as input data to optimize the setting parameters of the next product, which has the effect of parameter optimization. The products produced under the optimized setting parameters can better meet the quality requirements, which has the effect of optimizing product quality in the real-time production process.

[0104] For example, if the current production process is to compress by half, then the compression pressure will reflect the final compression effect to a certain extent. At this time, we can optimize our setting parameters based on the real-time parameter of compression pressure.

[0105] In summary, this embodiment has the following technical advantages:

[0106] 1. Accurate prediction: The DNN-based model can accurately predict the processing results, improve process stability and product quality.

[0107] 2. Real-time optimization: The system can monitor and adjust processing parameters in real time to reduce scrap rate and improve production efficiency.

[0108] 3. High degree of automation: The system integrates data collection, preprocessing, model training, prediction and optimization, and can run automatically to reduce human intervention.

[0109] Therefore, this embodiment has broad application prospects and significant economic benefits.

[0110] Embodiment 2

[0111] Please refer to Figure 4 A metal processing prediction and processing system 1 based on a deep neural network includes a memory 12, a processor 11, and a computer program stored in the memory 12 and executable on the processor 11, wherein the processor 11 implements the steps in the above-mentioned embodiment 1 when executing the computer program.

[0112] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0113] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0114] In the present invention, unless otherwise clearly specified and limited, when a first feature is “on” or “below” a second feature, it may be that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Moreover, when a first feature is “above”, “above” or “above” a second feature, it may be that the first feature is directly above or obliquely above the second feature, or it may simply mean that the first feature is higher in level than the second feature. When a first feature is “below”, “below” or “below” a second feature, it may be that the first feature is directly below or obliquely below the second feature, or it may simply mean that the first feature is lower in level than the second feature.

[0115] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0116] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A metal processing prediction and processing method based on a deep neural network, characterized in that: The following steps are involved: Obtaining various processing parameters and processing results in each metal processing process, and using the processing parameters, the processing results, the corresponding raw material parameters and the corresponding quality scoring parameters as metal processing data, wherein the processing parameters include equipment state parameters and equipment setting parameters, wherein the equipment state parameters refer to the state parameters of the metal processing equipment during the metal processing process, and the equipment setting parameters refer to the production parameters that need to be set in the metal processing equipment; Constructing a DNN model, inputting the metal processing data into the constructed DNN model for training, obtaining and saving the trained metal processing prediction model, using the raw material parameters, the equipment status parameters and the quality scoring parameters as input parameters of the DNN model, and selecting new equipment setting parameters as output parameters of the DNN model according to the processing results and the quality scoring parameters; Loading the metal processing prediction model, inputting the current raw material parameters, current equipment status parameters and current quality score parameters of the current product into the metal processing prediction model, and obtaining the equipment setting parameters to be optimized for the next product output by the metal processing prediction model; Before the production of the next product begins, the equipment setting parameters are adjusted according to the equipment setting parameters to be optimized.

2. The metal processing prediction and processing method based on deep neural network according to claim 1 is characterized in that: The selecting new equipment setting parameters as output parameters of the DNN model according to the processing results and the quality scoring parameters comprises the following steps: For a first input parameter, obtain a processing data set SQ that belongs to the same type of product and the same production line as the first input parameter and whose processing results reach the quality score parameter, and select a set of equipment setting parameters that best matches the first equipment setting parameters in the first metal processing data where the first input parameter is located as new equipment setting parameters from the processing data set SQ to obtain the output parameters of the first input parameter in the DNN model.

3. The metal processing prediction and processing method based on deep neural network according to claim 2 is characterized in that: The matching relationship between the two sets of device setting parameters is obtained by accumulating the product of the difference between each setting parameter in the two sets of device setting parameters and the corresponding weight value.

4. The metal processing prediction and processing method based on deep neural network according to claim 3 is characterized in that: The following steps are also included: If the amount of data in the processing data set SQ is lower than the lower limit threshold, the metal processing data of the same type of products and different production lines, and the metal processing data of the same production line but different types of products are added in turn, among which the conversion ratios of the metal processing data of the same type of products and the same production line, the metal processing data of the same type of products and different production lines, and the metal processing data of the same production line but different types of products are reduced in turn when matching equipment setting parameters.

5. The metal processing prediction and processing method based on deep neural network according to claim 1 is characterized in that: The next product is the next product of the same batch of the current product or the product of the next batch of the current product.

6. The metal processing prediction and processing method based on deep neural network according to any one of claims 1 to 5, characterized in that: The steps of constructing the DNN model include: Performing data preprocessing on the metal processing data, wherein the data preprocessing includes standardization processing, missing data filling and abnormal data processing; Before inputting the current raw material parameters, current equipment status parameters and current quality score parameters of the current product as input parameters into the metal processing prediction model, the following steps are also included: The current raw material parameters, current equipment status parameters and current quality score parameters of the current product are preprocessed.

7. The metal processing prediction and processing method based on deep neural network according to claim 6 is characterized in that: The data preprocessing is implemented using Python's Pandas library.

8. The metal processing prediction and processing method based on deep neural network according to any one of claims 1 to 5, characterized in that: The construction of the DNN model also includes the following steps: Define the hyperparameter search space and use Bayesian optimization methods to tune the hyperparameters of the DNN model.

9. The metal processing prediction and processing method based on deep neural network according to any one of claims 1 to 5, characterized in that: The equipment status parameters include the temperature of the equipment core components, the main power voltage, the core roller vibration amplitude and the air supply pressure; The raw material parameters include the raw material appearance size, the alloy ratio inside the raw material, the raw material initial temperature, and the material surface oil film thickness; The quality scoring parameters include product requirement parameters in the customer order, product surface flatness, average thickness and number of defects; The equipment setting parameters include the target temperature value for preheating the incoming materials, the set pressure under the equipment core and the advancing distance value for the incoming materials.

10. A metal processing prediction and processing system based on a deep neural network, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the metal processing prediction and processing method based on deep neural network described in any one of claims 1 to 9 is implemented.

Images