A calculation method for the structural strength of an injection molding gear mold based on data-driven design

Abstract

The present invention relates to the technical field of mold design, and specifically to a method for calculating the structural strength of an injection molded gear mold based on digital design. The method first obtains fixed parameter data, injection molding real-time data, and historical data; at the same time, finite element analysis is performed on the fixed parameter data to obtain first stress distribution data; then, the fixed parameter data and the injection molding real-time data are input into the trained multi-task local stress distribution prediction model to obtain a local area mask and local stress distribution prediction data; next, according to the real-time stress distribution influence weight output by the stress distribution influence prediction model, a stress distribution influence coefficient is obtained; according to the stress distribution influence coefficient and the first stress distribution data, second stress distribution data is obtained; by combining the local stress distribution prediction data, the local area mask, and the second stress distribution data, final stress distribution data is obtained; and the structural strength of the injection molded gear mold is obtained according to the final stress distribution data.

Description

Technical Field

[0001] The present invention relates to the technical field of mold design, and specifically provides a method for calculating the structural strength of an injection molded gear mold based on digital design. Background Art

[0002] With the development of modern manufacturing towards high precision, high efficiency and low cost, the demand for gears is increasing day by day, and the requirements for quality and performance are also constantly improving. Injection molded gears, with their advantages of high production efficiency, low cost and strong designability, gradually replace traditional metal gears in many fields, thus promoting the wide application of injection molded gear molds. In order to ensure the production quality of products, guarantee the service performance of molds and meet the safety requirements of production, it is essential to calculate the structural strength of injection molded gear molds.

[0003] At present, there are still deficiencies in the calculation of the structural strength of injection molded gear molds in the prior art; on the one hand, there are differences between the three-dimensional model of the modeled gear mold in the prior art and the actual mold, and it is impossible to accurately analyze the stress conditions in local areas such as chamfers, fillets and process holes; on the other hand, the prior art obtains the stress information of the gear mold during the injection process by manually inputting preset process parameters and combining with the finite element analysis method, but this cannot cope with the dynamic changes of parameters during the injection process, thus reducing the accuracy and reliability of the calculation of the structural strength of injection molded gear molds.

[0004] Therefore, a method for calculating the structural strength of an injection molded gear mold based on digital design is proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for calculating the structural strength of an injection molded gear mold based on digital design. This method first obtains fixed parameter data, real-time injection data and historical data; at the same time, finite element analysis is performed on the fixed parameter data to obtain the first stress distribution data; then, the fixed parameter data and real-time injection data are input into the trained multi-task local stress distribution prediction model to obtain the local area mask and local stress distribution prediction data; then, according to the real-time stress distribution influence weight output by the stress distribution influence prediction model, the stress distribution influence coefficient is obtained; according to the stress distribution influence coefficient and the first stress distribution data, the second stress distribution data is obtained; combining the local stress distribution prediction data, the local area mask and the second stress distribution data to obtain the final stress distribution data; the structural strength of the injection molded gear mold is obtained according to the final stress distribution data.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for calculating the structural strength of an injection molded gear mold based on digital design, comprising:

[0008] Obtain fixed parameter data, injection molding real-time data, and historical data;

[0009] Perform finite element analysis on the fixed parameter data to obtain first stress distribution data;

[0010] Construct a multi-task local stress distribution prediction model, train the multi-task local stress distribution prediction model using the historical data, input the fixed parameter data and the injection molding real-time data into the trained multi-task local stress distribution prediction model to obtain a local area mask and local stress distribution prediction data;

[0011] Construct a stress distribution influence prediction model, train the stress distribution influence prediction model using the historical data to obtain historical stress distribution influence weights; input the injection molding real-time data and the historical stress distribution influence weights into the trained stress distribution influence prediction model to obtain real-time stress distribution influence weights; obtain a stress distribution influence coefficient according to the real-time stress distribution influence weights;

[0012] Obtain second stress distribution data according to the stress distribution influence coefficient and the first stress distribution data; combine the local stress distribution prediction data, the local area mask, and the second stress distribution data to obtain final stress distribution data;

[0013] Obtain the structural strength of the injection molding gear mold according to the final stress distribution data.

[0014] Further, the fixed parameter data includes mold structure parameters and injection molding process parameters; the injection molding real-time data includes injection molding sensing data and injection molding stress data; the historical data includes historical parameter data, historical injection molding data, and historical stress distribution data.

[0015] Further, the process of training the multi-task local stress distribution prediction model using the historical data and inputting the fixed parameter data, the injection molding real-time data, and the first stress distribution data into the trained multi-task local stress distribution prediction model to obtain the local area mask and the local stress distribution prediction data includes:

[0016] Obtain historical parameter data, historical injection molding data, and historical stress distribution data from the historical data;

[0017] Input the historical parameter data, the historical injection molding data, and the historical stress distribution data into the multi-task local stress distribution prediction model for training to obtain a pre-trained local stress distribution prediction model;

[0018] Input the fixed parameter data and the real-time injection molding data into the pre-trained local stress distribution prediction model to obtain the local area mask and the stress distribution prediction data;

[0019] Obtain the local stress distribution prediction data according to the local area mask and the stress distribution prediction data.

[0020] Further, the multi-task local stress distribution prediction model includes: an input layer, a feature extraction layer, a local feature recognition layer, a stress distribution prediction layer based on the local area mask, and an output layer;

[0021] The input layer is used to perform feature transformation on the labeled input data to obtain data features;

[0022] The feature extraction layer is used to extract features from the data features to obtain deep features;

[0023] The local feature recognition layer is used to recognize the deep features to obtain local area mask features;

[0024] The stress distribution prediction layer based on the local area mask is used to obtain stress distribution prediction features according to the local area mask features and the data features;

[0025] The output layer is used to receive the local area mask features and the stress distribution prediction features to obtain the local area mask and the stress distribution prediction data.

[0026] Further, the process of training the stress distribution influence prediction model using the historical data to obtain the historical stress distribution influence weight, inputting the real-time injection molding data and the historical stress distribution influence weight into the trained stress distribution influence prediction model to obtain the real-time stress distribution influence weight, and obtaining the stress distribution influence coefficient according to the real-time stress distribution influence weight includes:

[0027] Obtain the real-time injection molding data;

[0028] Obtain historical injection molding data and historical stress distribution data from the historical data;

[0029] Input the historical injection molding data and the historical stress distribution data into the stress distribution influence prediction model for training to obtain the historical stress distribution influence weight;

[0030] Input the real-time injection molding data and the historical stress distribution influence weight into the trained stress distribution influence prediction model for processing to obtain the real-time stress distribution influence weight;

[0031] Perform a weighted sum of the weights of the influence of the real-time stress distribution on different injection molding data to obtain the stress distribution influence coefficient.

[0032] Further, the calculation formula for the stress distribution influence coefficient is:

[0033] ;

[0034] Wherein, represents the stress distribution influence coefficient; represents the number of types of injection molding data; represents the weighting factor of the i-th type of injection molding data; represents the weight of the influence of the real-time stress distribution of the i-th type of injection molding data.

[0035] Further, the stress distribution influence prediction model includes: an input layer, an encoding layer, a stress distribution influence prediction layer, a decoding layer, and an output layer;

[0036] The input layer is used to perform feature transformation on the model input data to obtain input features;

[0037] The encoding layer is used to perform encoding processing on the input features to obtain encoded features;

[0038] The stress distribution influence prediction layer is used to process the encoded features by adopting a ConvLSTM layer to obtain prediction features;

[0039] The decoding layer is used to perform decoding processing on the prediction features to obtain decoded prediction features;

[0040] The output layer is used to receive the decoded prediction features and process them by using a fully connected layer to obtain the stress distribution influence prediction result.

[0041] Further, the process of obtaining the second stress distribution data according to the stress distribution influence coefficient and the first stress distribution data; and combining the local stress distribution prediction data, the local area mask, and the second stress distribution data to obtain the final stress distribution data includes:

[0042] Obtain the stress distribution influence coefficient, the first stress distribution data, the local stress distribution prediction data, the local area mask, and the second stress distribution data;

[0043] Multiply the stress distribution influence coefficient and the first stress distribution data to obtain the second stress distribution data;

[0044] According to the local area mask and the second stress distribution data, obtain the second stress distribution data except for the local area;

[0045] Combine the second stress distribution data except for the local area with the local stress distribution prediction data to obtain the final stress distribution data.

[0046] Further, the calculation formula for obtaining the final stress distribution data is:

[0047] ;

[0048] Wherein, represents the final stress distribution data; represents the local stress distribution prediction data; represents the local area mask; represents the stress distribution influence coefficient; represents the element-wise product operation; represents the first stress distribution data.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] 1. The present invention proposes a multi-task local stress distribution prediction model for predicting the stress distribution data of a local area; this model uses a local feature recognition layer and a stress distribution prediction layer to respectively implement two tasks of regional feature recognition detection and stress distribution prediction; this model uses the local area mask output by the local feature recognition layer to make the stress distribution prediction layer focus on the stress distribution prediction of a specific area, thereby effectively analyzing and processing the stress distribution of local areas such as chamfers, fillets, and process holes, and further improving the accuracy and reliability of the structural strength calculation of an injection molded gear mold.

[0051] 2. The present invention proposes a stress distribution influence coefficient prediction method for dynamically adjusting stress distribution data; this method obtains a stress distribution influence coefficient according to the real-time stress distribution influence weight output by a stress distribution influence prediction model; the stress distribution can be adjusted by using the stress distribution influence coefficient during the injection molding process, thereby improving the accuracy and reliability of the structural strength calculation of an injection molded gear mold.

[0052] 3. The present invention proposes a stress distribution data calculation method for obtaining the final mold stress distribution data; this method first applies a stress distribution influence coefficient to the first stress distribution data to obtain an adjusted second stress distribution data; then combines the local stress distribution prediction data and the local area mask to obtain the final stress distribution data; the dynamically adjusted stress distribution data and the local stress distribution prediction data are fused through the calculation formula of the final stress distribution data, thereby improving the accuracy and reliability of the structural strength calculation of an injection molded gear mold. Description of the Drawings

[0053] Figure 1 Schematic flow chart of a method for calculating the structural strength of an injection molded gear mold based on digital design according to the present invention;

[0054] Figure 2 Schematic structural diagram of a multi-task local stress distribution prediction model according to the present invention. Specific embodiments

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with 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. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] The present invention proposes a method for calculating the structural strength of an injection molded gear mold based on digital design. The method proposed by the present invention will be specifically described below in conjunction with Embodiment 1 and Embodiment 2 as follows:

[0057] Embodiment 1

[0058] In order to improve the production quality and production strength of injection molded gears, a certain injection molded gear manufacturer uses a method for calculating the structural strength of an injection molded gear mold based on digital design proposed by the present invention. The flow of this method is shown as Figure 1 shown, and the specific process includes:

[0059] Obtain fixed parameter data, injection molding real-time data, and historical data;

[0060] Furthermore, the fixed parameter data includes mold structure parameters and injection molding process parameters; the injection molding real-time data includes injection molding sensor data and injection molding stress data; the historical data includes historical parameter data, historical injection molding data, and historical stress distribution data;

[0061] Furthermore, the mold structure parameters include: parting surface design parameters, cavity parameters, gating system parameters, cooling system parameters, etc.; the injection molding process parameters include: temperature parameters, pressure parameters, time parameters, etc.;

[0062] Furthermore, the injection molding sensor data includes injection molding temperature sensor data, injection molding pressure sensor data, injection molding flow rate sensor data, etc.; the injection molding stress data is discrete data collected by traditional methods such as strain gauge measurement method and ultrasonic detection method;

[0063] Furthermore, select some injection molding sensor data for numerical illustration, and specific reference can be made to Table 1;

[0064] Table 1 Numerical illustration of injection molding sensor data

[0065]

[0066] Further, the historical data is used to train the model and can be directly called from the historical database; the historical parameter data includes historical die structure parameters and historical injection molding process parameters; the historical injection molding data includes historical injection molding sensing data and historical injection molding stress data.

[0067] By introducing fixed parameter data, injection molding real-time data, and historical data, a reliable input data basis can be provided for the subsequent multi-task local stress distribution prediction model and stress distribution influence prediction model, thereby effectively improving the accuracy and reliability of the calculation of the structural strength of the injection molded gear die.

[0068] Perform finite element analysis on the fixed parameter data to obtain the first stress distribution data;

[0069] Further, the process of obtaining the first stress distribution data is as follows:

[0070] Establish a three-dimensional model based on the die structure parameters in the fixed parameter data;

[0071] Import the established model and material properties into the finite element analysis software;

[0072] Load the injection molding process parameters in the fixed parameter data and set the boundary conditions;

[0073] Select a suitable solver according to the model and start the solution program of the finite element analysis software to obtain the first stress distribution data.

[0074] The first stress distribution data obtained by finite element analysis in this embodiment is limited by the accuracy of model establishment and cannot accurately represent the stress distribution in local areas such as chamfers, fillets, and process holes.

[0075] Construct a multi-task local stress distribution prediction model, train the multi-task local stress distribution prediction model using historical data, input the fixed parameter data and injection molding real-time data into the trained multi-task local stress distribution prediction model, and obtain the local area mask and local stress distribution prediction data;

[0076] Further, the process of training the multi-task local stress distribution prediction model using historical data and inputting the fixed parameter data, injection molding real-time data, and the first stress distribution data into the trained multi-task local stress distribution prediction model to obtain the local area mask and local stress distribution prediction data includes:

[0077] Obtain historical parameter data, historical injection molding data, and historical stress distribution data from the historical data;

[0078] Input historical parameter data, historical injection molding data, and historical stress distribution data into a multi-task local stress distribution prediction model for training to obtain a pre-trained local stress distribution prediction model;

[0079] Input fixed parameter data and real-time injection molding data into the pre-trained local stress distribution prediction model to obtain a local area mask and stress distribution prediction data;

[0080] Based on the local area mask and stress distribution prediction data, obtain the local stress distribution prediction data.

[0081] In this embodiment, the two tasks of regional feature recognition detection and stress distribution prediction are respectively realized by using the local feature recognition layer and the stress distribution prediction layer in the multi-task local stress distribution prediction model; the model uses the local area mask output by the local feature recognition layer to enable the stress distribution prediction layer to focus on the stress distribution prediction of specific regions, thereby effectively analyzing and processing the stress distribution of local regions such as chamfers, fillets, and process holes, and further improving the accuracy and reliability of the structural strength calculation of injection molding gear dies.

[0082] Furthermore, the structure of the multi-task local stress distribution prediction model can refer to Figure 2 This model includes: an input layer, a feature extraction layer, a local feature recognition layer, a stress distribution prediction layer based on a local area mask, and an output layer;

[0083] The input layer is used to perform feature transformation on the labeled input data to obtain data features;

[0084] Furthermore, the input data includes historical parameter data and historical injection molding data; the historical stress distribution data is used for comparison of the model output.

[0085] The feature extraction layer is used to extract features from the data features to obtain deep features;

[0086] The local feature recognition layer is used to recognize the deep features to obtain local area mask features;

[0087] Furthermore, the local feature recognition layer adopts a residual convolutional network, which includes multiple residual blocks and a fully connected layer; the residual blocks learn the mapping relationship between the input features and the output features through skip connections and are used to extract the feature relationships between different regions; the fully connected layer is used to output the local area mask features.

[0088] The stress distribution prediction layer based on the local area mask is used to obtain stress distribution prediction features according to the local area mask features and the data features;

[0089] Furthermore, the stress distribution prediction layer based on the local region mask includes a mask modulation branch and an LSTM branch. The mask modulation branch uses 4 serial mask modulation residual blocks to learn the spatial features of the stress distribution within the local region; the LSTM branch uses 4 serial LSTM layers to learn the temporal variation features of the stress distribution; finally, the stress distribution prediction layer based on the local region mask fuses the output features of different branches through a channel merging operation to obtain the stress distribution prediction features;

[0090] Furthermore, the expression of the mask modulation residual block is as follows:

[0091] ;

[0092] where, represents the output feature of the mask modulation residual block; represents the Relu activation function; represents a 3×3 convolutional kernel with independent parameters; represents the local mask modulation layer; represents the input feature of the mask modulation residual block; represents the local region mask feature.

[0093] Furthermore, the expression of the local mask modulation layer is as follows:

[0094] ;

[0095] where, represents the output feature of the local mask modulation layer; represents a 1×1 convolutional kernel with independent parameters; represents the local region mask feature; represents the element-wise product operation; represents the input feature of the local mask modulation layer.

[0096] The output layer is used to receive the local region mask feature and the stress distribution prediction feature to obtain the local region mask and the stress distribution prediction data.

[0097] In this embodiment, the multi-task local stress distribution prediction model provides accurate local mask information for the stress distribution prediction layer based on the local region mask by inputting the extracted features into the local feature recognition layer, and then uses different branches in the stress distribution prediction layer and combines mask feature modulation, enabling the model to focus on the stress distribution prediction within a specific region, thereby improving the accuracy of the final stress distribution data calculation.

[0098] Build a prediction model for the influence of the build stress distribution, and use historical data to train the prediction model for the influence of the stress distribution to obtain the historical weight of the influence of the stress distribution; input the real-time injection molding data and the historical weight of the influence of the stress distribution into the trained prediction model for the influence of the stress distribution to obtain the real-time weight of the influence of the stress distribution; obtain the influence coefficient of the stress distribution according to the real-time weight of the influence of the stress distribution;

[0099] Further, the process of using historical data to train the prediction model for the influence of the stress distribution to obtain the historical weight of the influence of the stress distribution, inputting the real-time injection molding data and the historical weight of the influence of the stress distribution into the trained prediction model for the influence of the stress distribution to obtain the real-time weight of the influence of the stress distribution, and obtaining the influence coefficient of the stress distribution according to the real-time weight of the influence of the stress distribution includes:

[0100] Obtain the real-time injection molding data;

[0101] Obtain the historical injection molding data and the historical stress distribution data from the historical data;

[0102] Input the historical injection molding data and the historical stress distribution data into the prediction model for the influence of the stress distribution for training to obtain the historical weight of the influence of the stress distribution;

[0103] Input the real-time injection molding data and the historical weight of the influence of the stress distribution into the trained prediction model for the influence of the stress distribution for processing to obtain the real-time weight of the influence of the stress distribution;

[0104] Perform weighted summation on the real-time weights of the influence of the stress distribution for different injection molding data to obtain the influence coefficient of the stress distribution.

[0105] Obtain the influence coefficient of the stress distribution by using the real-time weight of the influence of the stress distribution output by the prediction model for the influence of the stress distribution; the dynamic parameters in the injection molding process can be used to adjust the stress distribution by using the influence coefficient, thereby improving the accuracy and reliability of the calculation of the structural strength of the injection molding gear die.

[0106] Further, the calculation formula for the influence coefficient of the stress distribution is:

[0107] ;

[0108] Among them, represents the influence coefficient of the stress distribution; represents the number of types of injection molding data; represents the weighting factor of the i-th type of injection molding data; represents the real-time weight of the influence of the stress distribution of the i-th type of injection molding data.

[0109] Furthermore, the real-time stress distribution influence weight, stress distribution influence coefficient, and stress distribution data are spatially distributed, and the weight, coefficient, and data corresponding to each regional position in the mold are in one-to-one correspondence; the stress distribution influence weights that each regional position may be affected by different injection molding data are all different.

[0110] Furthermore, the number of types of injection molding data and the weighting factors corresponding to the injection molding data can be flexibly set by those skilled in the art.

[0111] By weighting the injection molding data influence weights that may affect the stress distribution using the weighting factor, the comprehensive real-time stress distribution influence weight, that is, the stress distribution influence coefficient, is obtained; since the data collected during the injection molding process is dynamically changing, and the degree of influence on the stress distribution is also dynamically changing, the calculated stress distribution influence coefficient is also changing; the stress distribution influence coefficient can be used to adjust and correct the stress distribution data simulated by the software, thereby improving the accuracy and reliability of the calculation of the structural strength of the injection molded gear mold.

[0112] Furthermore, the stress distribution influence prediction model includes: an input layer, an encoding layer, a stress distribution influence prediction layer, a decoding layer, and an output layer;

[0113] The input layer is used to perform feature transformation on the model input data to obtain input features;

[0114] The encoding layer is used to perform encoding processing on the input features to obtain encoded features;

[0115] The stress distribution influence prediction layer is used to process the encoded features by adopting a ConvLSTM layer to obtain prediction features;

[0116] The decoding layer is used to perform decoding processing on the prediction features to obtain decoded prediction features;

[0117] The output layer is used to receive the decoded prediction features and process them using a fully connected layer to obtain the stress distribution influence prediction result.

[0118] By utilizing the ability of the ConvLSTM layer in the stress distribution influence prediction model to extract time series features to predict the trend of stress distribution changes, combined with the advantages of the deep learning model, the stress distribution influence weights under different injection molding processes can be accurately obtained, and this weight can be used to effectively adjust the stress distribution data in real time.

[0119] According to the stress distribution influence coefficient and the first stress distribution data, the second stress distribution data is obtained; combining the local stress distribution prediction data, the local area mask, and the second stress distribution data, the final stress distribution data is obtained;

[0120] Further, based on the stress distribution influence coefficient and the first stress distribution data, the second stress distribution data is obtained; the implementation process of obtaining the final stress distribution data by combining the local stress distribution prediction data, the local area mask, and the second stress distribution data includes:

[0121] Obtain the stress distribution influence coefficient, the first stress distribution data, the local stress distribution prediction data, the local area mask, and the second stress distribution data;

[0122] Multiply the stress distribution influence coefficient and the first stress distribution data to obtain the second stress distribution data;

[0123] According to the local area mask and the second stress distribution data, obtain the second stress distribution data outside the local area;

[0124] Combine the second stress distribution data outside the local area with the local stress distribution prediction data to obtain the final stress distribution data.

[0125] By combining the dynamically adjusted stress distribution data, the local area mask, and the local stress distribution prediction data, comprehensive and accurate die stress distribution data can be obtained, thereby improving the accuracy and reliability of the structural strength calculation of the injection molded gear die.

[0126] Further, the calculation formula for obtaining the final stress distribution data is:

[0127] ;

[0128] Wherein, represents the final stress distribution data; represents the local stress distribution prediction data; represents the local area mask; represents the stress distribution influence coefficient; represents the element-wise product operation; represents the first stress distribution data.

[0129] Further, the value range of the local area mask is [0, 1]; represents the non-local area mask.

[0130] In this embodiment, the calculation formula of the final stress distribution data can organically combine the output data of the multi-task local stress distribution prediction model and the finite element analysis data adjusted by the influence coefficient through the local area mask, thereby effectively improving the accuracy and reliability of the structural strength calculation of the injection molded gear die.

[0131] According to the final stress distribution data, obtain the structural strength of the injection molded gear die.

[0132] This embodiment provides a method for calculating the structural strength of an injection molded gear mold based on digital design. The method first obtains fixed parameter data, real-time injection molding data, and historical data; then, the fixed parameter data and real-time injection molding data are input into the trained multi-task local stress distribution prediction model to obtain the local area mask and local stress distribution prediction data; next, according to the real-time stress distribution influence weight output by the stress distribution influence prediction model, the stress distribution influence coefficient is obtained; according to the stress distribution influence coefficient and the first stress distribution data obtained by finite element analysis, the second stress distribution data is obtained; combining the local stress distribution prediction data, the local area mask, and the second stress distribution data to obtain the final stress distribution data; the structural strength of the injection molded gear mold is obtained according to the final stress distribution data. This method can effectively improve the accuracy and reliability of calculating the structural strength of the injection molded gear mold.

[0133] Embodiment 2

[0134] The present invention proposes a method for calculating the structural strength of an injection molded gear mold based on digital design. In order to further verify the effectiveness of the model and method flow proposed by the present invention, the present invention conducts model accuracy tests and method flow accuracy tests for different model selections and different method flows respectively. The present invention selects two injection molded gear manufacturers, A and B, to conduct the above two groups of accuracy tests respectively.

[0135] The present invention selects the historical data of manufacturer A of injection molded gears in the past 3 years as the dataset of the model, where the data of the first and second years are used as the training set of the model, and the data of the third year are used as the validation set; the sampling of the dataset of the model refers to the following rules: taking a week as a unit, randomly selecting 5 days of data per week; among them, 2 groups of data are selected in the morning, noon, and evening every day; all data come from the same type of gear mold.

[0136] Furthermore, the historical data includes: historical parameter data, historical injection molding data, and historical stress distribution data.

[0137] The present invention inputs the training sets collected from manufacturer A of injection molded gears into different models for training to obtain their respective pre-trained models; then, the validation sets are input into each pre-trained model to obtain model prediction data; then, the model prediction data is compared with the real data for testing to obtain the model accuracy test results; the model accuracy tests are respectively recorded as Test 1, Test 2, Test 3, and Test 4. The models corresponding to each test are: the multi-task local stress distribution prediction model proposed by the present invention, denoted as Model 1; the local stress distribution prediction model without a local feature recognition layer and the local area mask is obtained by manual adjustment, denoted as Model 2; the local stress distribution prediction model without a local feature recognition layer and without a local area mask, denoted as Model 3; the local stress distribution prediction model using traditional CNN, denoted as Model 4.

[0138] The present invention compares the model prediction data with the real data using a preset threshold, and obtains the accuracy test results under different models according to the proportion within a reasonable range; the model accuracy test results are shown in Table 2.

[0139] Table 2 Model Accuracy Test Results

[0140]

[0141] It can be seen from the results in Table 2 that the model proposed by the present invention has better test results in terms of accuracy test results than those of other models; thus, it can be shown that the multi-task local stress distribution prediction model proposed by the present invention can predict accurate local stress distribution data, and further improve the accuracy of the structural strength of the injection molded gear mold;

[0142] In order to further test the accuracy of the method flow, this embodiment collects the historical data of B injection molded gear manufacturers in the past two years, according to the same data sampling rules, and then divides the sampled data set into 4 groups of data for testing, which are respectively recorded as: test sample one, test sample two, test sample three, and test sample four; the test data of each group is processed according to method flow one and method flow two respectively to obtain the final stress distribution data one and the final stress distribution data two; among them, method flow one is the method proposed by the present invention for calculating the final stress distribution data; method flow two removes the process of adjusting the influence coefficient of the first stress distribution data in the method flow of the present invention; finally, combined with the verification of the historical real stress distribution data, the test accuracy rates under different processes are obtained; the method flow accuracy test results are shown in Table 3;

[0143] Table 3 Method Flow Accuracy Test Results

[0144]

[0145] It can be seen from the results in Table 3 that the accuracy test results obtained by using method flow one, that is, the scheme proposed by the present invention for calculating the final stress distribution data, are better than those obtained by using method flow two, which shows that it is necessary to dynamically adjust the first stress distribution data using the stress distribution influence coefficient, and can improve the real-time accuracy of the structural strength of the injection molded gear mold.

[0146] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirits of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for calculating the structural strength of an injection gear mold based on data-based design, characterized in that: include: Obtain fixed parameter data, injection molding real-time data and historical data; Using finite element analysis on the fixed parameter data to obtain first stress distribution data; Constructing a multi-task local stress distribution prediction model, using the historical data to train the multi-task local stress distribution prediction model, inputting the fixed parameter data and the injection molding real-time data into the trained multi-task local stress distribution prediction model, and obtaining a local area mask and local stress distribution prediction data; Constructing a stress distribution influence prediction model, using the historical data to train the stress distribution influence prediction model, and obtaining the historical stress distribution influence weight; inputting the injection molding real-time data and the historical stress distribution influence weight into the trained stress distribution influence prediction model, and obtaining the real-time stress distribution influence weight; and obtaining the stress distribution influence coefficient according to the real-time stress distribution influence weight; Obtaining second stress distribution data according to the stress distribution influence coefficient and the first stress distribution data; and obtaining final stress distribution data by combining the local stress distribution prediction data, the local area mask and the second stress distribution data; The process of realizing the final stress distribution data includes: Acquire the stress distribution influence coefficient, the first stress distribution data, the local stress distribution prediction data, the local area mask and the second stress distribution data; Multiplying the stress distribution influence coefficient and the first stress distribution data to obtain the second stress distribution data; Obtaining the second stress distribution data except for the local area according to the local area mask and the second stress distribution data; Combining the second stress distribution data except for the local area with the local stress distribution prediction data to obtain the final stress distribution data; The calculation formula for obtaining the final stress distribution data is: ； in, represents the final stress distribution data; represents the local stress distribution prediction data; represents the local area mask; represents the stress distribution influence coefficient; Represents an element-wise product operation; represents the first stress distribution data; According to the final stress distribution data, the structural strength of the injection gear mold is obtained.

2. The method for calculating the structural strength of an injection gear mold based on digital design according to claim 1 is characterized in that: The fixed parameter data includes mold structure parameters and injection molding process parameters; the real-time injection molding data includes injection molding sensor data and injection molding stress data; and the historical data includes historical parameter data, historical injection molding data and historical stress distribution data.

3. The method for calculating the structural strength of an injection gear mold based on digital design according to claim 1 is characterized in that: The process of training the multi-task local stress distribution prediction model by using the historical data, inputting the fixed parameter data, the injection real-time data and the first stress distribution data into the trained multi-task local stress distribution prediction model, and obtaining the local area mask and the local stress distribution prediction data includes: Acquiring historical parameter data, historical injection molding data and historical stress distribution data from the historical data; Inputting the historical parameter data, the historical injection molding data and the historical stress distribution data into the multi-task local stress distribution prediction model for training to obtain a pre-trained local stress distribution prediction model; Inputting the fixed parameter data and the injection molding real-time data into the pre-trained local stress distribution prediction model to obtain the local area mask and the stress distribution prediction data; The local stress distribution prediction data is obtained according to the local area mask and the stress distribution prediction data.

4. The method for calculating the structural strength of an injection gear mold based on digital design according to claim 3 is characterized in that: The multi-task local stress distribution prediction model includes: an input layer, a feature extraction layer, a local feature recognition layer, a stress distribution prediction layer based on a local area mask, and an output layer; The input layer is used to perform feature transformation on the labeled input data to obtain data features; The feature extraction layer is used to extract the data features to obtain deep features; The local feature recognition layer is used to recognize the deep features and obtain local area mask features; The stress distribution prediction layer based on the local area mask is used to obtain stress distribution prediction features according to the local area mask features and the data features; The output layer is used to receive the local area mask features and the stress distribution prediction features to obtain the local area mask and the stress distribution prediction data.

5. The method for calculating the structural strength of an injection gear mold based on digital design according to claim 1 is characterized in that: The stress distribution influence prediction model is trained using the historical data to obtain the historical stress distribution influence weight; the injection molding real-time data and the historical stress distribution influence weight are input into the trained stress distribution influence prediction model to obtain the real-time stress distribution influence weight; The process of obtaining the stress distribution influence coefficient according to the real-time stress distribution influence weight includes: Acquiring the injection molding real-time data; Acquire historical injection molding data and historical stress distribution data from the historical data; Inputting the historical injection molding data and the historical stress distribution data into the stress distribution influence prediction model for training to obtain the historical stress distribution influence weight; Inputting the injection molding real-time data and the historical stress distribution influence weight into the trained stress distribution influence prediction model for processing to obtain the real-time stress distribution influence weight; The real-time stress distribution influence weights of different injection molding data are weighted summed to obtain the stress distribution influence coefficient.

6. The method for calculating the structural strength of an injection gear mold based on digital design according to claim 5 is characterized in that: The calculation formula of the stress distribution influence coefficient is: ； in, represents the stress distribution influence coefficient; Indicates the number of types of injection molding data; Indicates Weighting factors for injection molding data; Indicates The real-time stress distribution of the injection molding data affects the weight.

7. The method for calculating the structural strength of an injection gear mold based on digital design according to claim 1 is characterized in that: The stress distribution impact prediction model comprises: an input layer, a coding layer, a stress distribution impact prediction layer, a decoding layer and an output layer; The input layer is used to perform feature transformation on the model input data to obtain input features; The encoding layer is used to encode the input features to obtain encoded features; The stress distribution impact prediction layer is used to process the encoding features by using a ConvLSTM layer to obtain prediction features; The decoding layer is used to decode the prediction feature to obtain a decoded prediction feature; The output layer is used to receive the decoded prediction features and process them using a fully connected layer to obtain a stress distribution impact prediction result.

Images