Accurate batching system for casting aluminum alloy

Abstract

The invention discloses a cast aluminum alloy accurate batching system, which comprises a historical data collection module used for collecting historical data during aluminum alloy casting; the historical data processing module is used for cleaning and normalizing the collected historical data; the feature selection module is used for performing correlation analysis and feature importance evaluation; the model training module is used for training a prediction model of the ingredient proportion based on a CNN (convolutional neural network); the feedback control module is used for forming the prediction result into an output instruction; and the execution control module is used for realizing weighing of the ingredients and automatic control alarm according to the output instruction. A model training module is established to predict the ingredient proportion, a regression loss function is calculated based on a mean square error, the regression loss function is minimized through iterative training, and then the accuracy of the predicted ingredient proportion is guaranteed; and automatic batching and control are realized through the feedback control module and the execution control module.

Description

Technical Field

[0001] The invention relates to a batching system, in particular to a precise batching system for casting aluminum alloy. Background Art

[0002] Casting aluminum alloy batching refers to which materials should be added during the process of smelting a certain aluminum alloy so that the chemical elements of the aluminum alloy can reach the specified standards, and the total weight of the materials should be as close as possible to the weight of the molten aluminum alloy; the quality of aluminum alloy casting is affected by many factors, such as the refinement of the casting process, the quality of the casting equipment, the selection of raw materials and the batching scheme, etc. The design of the batching scheme is the key point in the field of aluminum alloy casting; the chemical elements of aluminum alloys generally include multiple elements, among which the main elements are Fe and Si, and other elements are generally trace elements. When adding materials to supplement the main elements, trace elements will be introduced at the same time, and when adding materials to supplement the trace elements, the main elements will be introduced at the same time. It is necessary to ensure that these elements do not exceed the fixed standards;

[0003] In the process of aluminum alloy batching, operators usually perform manual calculations and verifications based on established batching rules. However, due to the complexity of batching rules and multiple conditional restrictions, as well as the technical difficulty of recalculating when the verification is inaccurate, it is often difficult for operators to accurately calculate the batching value in a short period of time. This leads to the emergence of numerical deviations, which in turn affect the high-quality casting production of aluminum alloys and increase the time cost of industrial production. Summary of the invention

[0004] In order to solve the deficiencies of the above-mentioned technologies, the present invention provides a precise batching system for casting aluminum alloys.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a precise batching system for casting aluminum alloy, including a historical data collection module, which is used to collect historical data during aluminum alloy casting, and the historical data includes raw material composition data, batching ratio and production environment parameters;

[0006] The historical data processing module is used to clean and normalize the collected historical data to eliminate the influence of outliers; the normalization formula is:

[0007]

[0008] Among them, x' is the normalized data; x i is the value of the i-th historical data; x max is the maximum value in historical data; x min is the minimum value in historical data;

[0009] Feature selection module, which is used to select the features that have the greatest impact on the quality of aluminum alloy as model input through correlation analysis and feature importance evaluation;

[0010] Model training module, used to train the prediction model of ingredient ratio based on convolutional neural network (CNN);

[0011] The feedback control module is used to collect raw material composition and production environment parameters in real time, and according to the prediction of the proportion of ingredients, the prediction results are formed into output instructions, and the output instructions are sent to the execution control module;

[0012] The execution control module is used to control the batching scale and conveying equipment according to the output instructions and based on the PLC controller to achieve the weighing of the ingredients and automatic control alarm.

[0013] Furthermore, the calculation formula of the correlation coefficient r in the correlation analysis is:

[0014]

[0015] Among them, x i is the value of the i-th historical data; is the average value of the historical data.

[0016] Furthermore, the method for evaluating feature importance is to evaluate the importance of features based on a tree model, that is, to calculate the number of times each feature is used in the tree model or used for splitting based on the XGBoost model to measure the importance of the feature.

[0017] Furthermore, the feature importance is evaluated based on the XGBoost model, including the following steps:

[0018] Step 1: Build a data set based on historical data;

[0019] Step 2: Divide the dataset into features and target variables, and divide it into training set and test set;

[0020] Step 3: Initialize the XGBoost model and optimize the hyperparameters based on grid search;

[0021] Step 4: Find the optimal hyperparameters based on the five-fold cross-validation technique and evaluate the performance of the XGBoost model based on the negative mean square error.

[0022] Step 5: Get the contribution of each feature and sort the contribution in descending order;

[0023] Step 6: Sort the names of each feature and the corresponding contribution in descending order and store them;

[0024] Step 7: Based on the stored results, evaluate the feature importance.

[0025] Furthermore, in the model training module, the regression loss function is calculated based on the mean square error (MSE), and the Adam optimizer is used to update the model parameters. Through iterative training, the model parameters are adjusted to minimize the regression loss function.

[0026] The calculation formula for mean square error MSE is:

[0027]

[0028] Where n is the number of samples; x i is the actual ingredient ratio of the i-th sample; y i is the predicted ingredient ratio of the ith sample.

[0029] Furthermore, after training the prediction model, the model training module optimizes the prediction model based on the Dropout layer to prevent the prediction model from overfitting.

[0030] Further, the feedback control module includes a temperature sensor for monitoring temperature changes during the aluminum alloy smelting process;

[0031] The composition analyzer is used to analyze the chemical composition of aluminum alloy raw materials and ensure the accuracy of the ingredient ratio by accurately measuring the content of various elements.

[0032] Furthermore, the execution control module includes a weight sensor, which is installed on the batching scale to collect weight data of various raw materials in real time to ensure the accuracy of the batching.

[0033] The invention discloses a precise batching system for casting aluminum alloys. The system predicts the batching ratio by establishing a model training module, calculates the regression loss function based on the mean square error, and minimizes the regression loss function through iterative training, thereby ensuring the accuracy of the predicted batching ratio, reducing the numerical deviation, and improving the high-quality melting and casting productivity of the aluminum alloy. Automatic batching and control are achieved through a feedback control module and an execution control module, saving manpower. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a system architecture diagram of the present invention.

[0035] Figure 2 Schematic diagram of the process flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0037] Embodiment 1:

[0038] like Figure 1 and2 A casting aluminum alloy precise batching system shown includes a historical data collection module for collecting historical data during aluminum alloy casting, the historical data including raw material composition data, batching ratio and production environment parameters; the production environment parameters include production temperature and cooling speed;

[0039] The historical data processing module is used to clean and normalize the collected historical data to eliminate the influence of outliers; the normalization formula is:

[0040]

[0041] Among them, x' is the normalized data; x i is the value of the i-th historical data; x max is the maximum value in historical data; x min is the minimum value in historical data;

[0042] Normalization refers to normalizing the data column to a fixed interval. In order to make it easier to process data, the data is mapped to the range of 0 to 1.

[0043] The feature selection module is used to select the features that have the greatest impact on the quality of aluminum alloys as model inputs through correlation analysis and feature importance assessment. The purpose of correlation analysis is to evaluate the strength of the relationship between each feature and the quality of aluminum alloys. By calculating the correlation coefficient between the feature and the quality index, the features that are highly correlated with the quality index of aluminum alloys are identified. The value of the correlation coefficient is between -1 and 1. The closer the value is to 1 or -1, the stronger the correlation between the feature and the quality index of aluminum alloys.

[0044] The calculation formula of the correlation coefficient r in correlation analysis is:

[0045]

[0046] Among them, x i is the value of the i-th historical data; is the average value of the historical data;

[0047] Feature importance assessment is to evaluate the influence of features on the target variable, that is, the importance of features in the model. The importance of features is calculated based on the tree model, and the importance of each feature is calculated based on the XGBoost model to measure the importance of features by calculating the number of times each feature is used in the tree model or the importance of being used for splitting. In the XGBoost model, the more features are used to build decision trees, the higher their feature importance is; attribute importance is calculated by calculating and ranking each attribute in the data set; in a single decision tree, attribute importance is calculated by the amount of improvement in performance measurement at each attribute split point;

[0048] The method for evaluating feature importance is to evaluate the importance of features based on the tree model, that is, to calculate the number of times each feature is used in the tree model or the importance of being used for splitting based on the XGBoost model to measure the importance of the feature;

[0049] Evaluating feature importance based on the XGBoost model involves the following steps:

[0050] Step 1: Build a data set based on historical data;

[0051] Step 2: Divide the dataset into features and target variables, and divide it into training set and test set;

[0052] Step 3: Initialize the XGBoost model and optimize the hyperparameters based on grid search;

[0053] Step 4: Find the optimal hyperparameters based on the five-fold cross-validation technique and evaluate the performance of the XGBoost model based on the negative mean square error.

[0054] Step 5: Get the contribution of each feature and sort the contribution in descending order;

[0055] Step 6: Sort the names of each feature and the corresponding contribution in descending order and store them;

[0056] Step 7: Evaluate the importance of features based on the stored results;

[0057] Model training module, used to train the prediction model of ingredient ratio based on convolutional neural network (CNN);

[0058] In the model training module, the regression loss function is calculated based on the mean square error (MSE), and the Adam optimizer is used to update the model parameters. Through iterative training, the model parameters are adjusted to minimize the regression loss function.

[0059] The calculation formula for mean square error MSE is:

[0060]

[0061] Where n is the number of samples; x i is the actual ingredient ratio of the i-th sample; y i is the predicted ingredient ratio of the i-th sample;

[0062] After training the prediction model, the model training module optimizes the prediction model based on the Dropout layer to prevent the prediction model from overfitting;

[0063] The feedback control module is used to collect raw material composition and production environment parameters in real time, and according to the prediction of the proportion of ingredients, the prediction results are formed into output instructions, and the output instructions are sent to the execution control module;

[0064] The feedback control module includes a temperature sensor for monitoring temperature changes during the aluminum alloy melting process;

[0065] Composition analyzer, used to analyze the chemical composition of aluminum alloy raw materials, and ensure the accuracy of the proportion of ingredients by accurately measuring the content of various elements;

[0066] The execution control module is used to control the batching scale and conveying equipment based on the PLC controller according to the output instructions to achieve the weighing and automatic control alarm of the batching;

[0067] The execution control module includes a weight sensor, which is used to be installed on the batching scale to collect the weight data of various raw materials in real time to ensure the accuracy of the batching.

[0068] Embodiment 2:

[0069] Convolutional neural network (CNN) is a deep learning model or a multi-layer perceptron similar to an artificial neural network, including a data input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer. A prediction model for ingredient ratio is established based on the convolutional neural network (CNN), including the following architecture:

[0070] Through the data input layer, the pre-processed raw material composition data and production environment are used as the input of the prediction model; through the convolution layer, the local features in the input data are extracted through the convolution kernel, and the network captures the spatial relationship and pattern in the data; the convolution layer is followed by the activation function, namely the ReLU function, to perform nonlinear transformation, thereby enhancing the expression ability of the prediction model; through the average pooling layer in the pooling layer, the spatial size of the data is effectively reduced; after the extraction of the convolution layer and the dimensionality reduction processing of the pooling layer, the fully connected layer globally integrates the local features to generate the final ingredient ratio prediction result, that is, the input feature vector is multiplied by the weight matrix, and the bias parameter is added to output the accurate ingredient ratio; the formula of the fully connected layer is:

[0071] y = σ(ω*x+b);

[0072] Among them, y is the output ingredient ratio vector; σ is the ReLU activation function; x is the input vector; ω is the weight matrix; and b is the bias parameter.

[0073] The present invention can accurately predict the proportion of ingredients by constructing a model training module. This process involves calculation based on mean square error, thereby obtaining a regression loss function. Through continuous iterative training, we are committed to minimizing this regression loss function to ensure that the predicted proportion of ingredients has a high degree of accuracy. This method significantly reduces numerical deviations, thereby improving the efficiency of high-quality aluminum alloy casting production.

[0074] In addition, the present invention introduces a feedback control module and an execution control module, which work together to realize the automation of the batching process. In this way, we can not only accurately control the batching ratio, but also monitor and adjust the production process in real time to ensure the stability and consistency of the production process. This automated control system greatly saves human resources, reduces errors and uncertainties in human operations, and further improves production efficiency and product quality.

[0075] The above implementation modes are not limitations of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the scope of the technical solution of the present invention also belong to the protection scope of the present invention.

Claims

1. A precise batching system for casting aluminum alloy, characterized in that: It includes a historical data collection module for collecting historical data of aluminum alloy casting, the historical data including raw material composition data, ingredient ratio and production environment parameters; The historical data processing module is used to clean and normalize the collected historical data to eliminate the influence of outliers; the normalization formula is: Among them, x' is the normalized data; x i is the value of the i-th historical data; x max is the maximum value in historical data; x min is the minimum value in historical data; Feature selection module, which is used to select the features that have the greatest impact on the quality of aluminum alloy as model input through correlation analysis and feature importance evaluation; Model training module, used to train the prediction model of ingredient ratio based on convolutional neural network (CNN); The feedback control module is used to collect raw material composition and production environment parameters in real time, and according to the prediction of the proportion of ingredients, the prediction results are formed into output instructions, and the output instructions are sent to the execution control module; The execution control module is used to control the batching scale and conveying equipment according to the output instructions and based on the PLC controller to achieve the weighing of the ingredients and automatic control alarm.

2. The precise batching system for casting aluminum alloy according to claim 1 is characterized in that: The calculation formula of the correlation coefficient r in the correlation analysis is: Among them, x i is the value of the i-th historical data; is the average value of the historical data.

3. The precise batching system for casting aluminum alloy according to claim 1 is characterized in that: The method for evaluating the importance of features is to evaluate the importance of features based on a tree model, that is, to calculate the number of times each feature is used in the tree model or the importance of being used for splitting based on the XGBoost model to measure the importance of the feature.

4. The precise batching system for casting aluminum alloy according to claim 3 is characterized in that: The feature importance evaluation based on the XGBoost model includes the following steps: Step 1: Build a data set based on historical data; Step 2: Divide the dataset into features and target variables, and divide it into training set and test set; Step 3: Initialize the XGBoost model and optimize the hyperparameters based on grid search; Step 4: Find the optimal hyperparameters based on the five-fold cross-validation technique and evaluate the performance of the XGBoost model based on the negative mean square error. Step 5: Get the contribution of each feature and sort the contribution in descending order; Step 6: Sort the names of each feature and the corresponding contribution in descending order and store them; Step 7: Based on the stored results, evaluate the feature importance.

5. The precise batching system for casting aluminum alloy according to claim 1, characterized in that: In the model training module, the regression loss function is calculated based on the mean square error (MSE), and the model parameters are updated using the Adam optimizer. The model parameters are adjusted through iterative training to minimize the regression loss function. The calculation formula for mean square error MSE is: Where n is the number of samples; x i is the actual ingredient ratio of the i-th sample; y i is the predicted ingredient ratio of the ith sample.

6. The precise batching system for casting aluminum alloy according to claim 1, characterized in that: After training the prediction model, the model training module optimizes the prediction model based on the Dropout layer to prevent the prediction model from overfitting.

7. The precise batching system for casting aluminum alloy according to claim 1, characterized in that: The feedback control module includes a temperature sensor for monitoring temperature changes during the aluminum alloy smelting process; The composition analyzer is used to analyze the chemical composition of aluminum alloy raw materials and ensure the accuracy of the ingredient ratio by accurately measuring the content of various elements.

8. The precise batching system for casting aluminum alloy according to claim 1, characterized in that: The execution control module includes a weight sensor, which is installed on the batching scale to collect weight data of various raw materials in real time to ensure the accuracy of the batching.

Images