AI-based injection mold rapid design method and device

Abstract

The invention is suitable for the technical field of injection mold design, and provides an AI-based injection mold rapid design method and device, and the method is applied to an AI-based injection mold design control system. The system comprises an information acquisition module, a model construction module, a model training module, a model evaluation module, a parameter determination module, a scheme determination module, a wireless communication module, a memory, an alarm, a processing center and an intelligent mobile terminal. The information acquisition module, the model construction module, the model training module, the model evaluation module, the parameter determination module, the scheme determination module and the wireless communication module are respectively connected with the processing center; the invention further provides an AI-based injection mold design control device. The mold process parameters can be preset for design, and various forming quality problems are prevented in advance.

Description

Technical Field

[0001] The present invention relates to the technical field of injection mold design, and specifically to a method and device for rapid design of injection molds based on AI in the field of artificial intelligence. Background Art

[0002] Since an injection mold is a production tool for injection molded parts, the quality of the injection mold design is ultimately determined by detecting the quality of the injection molded parts through injection molding. In the current process of injection mold design, from receiving an order to the mold trial production without abnormalities until normal mass production, currently, according to the customer's product drawings, the general parts are designed by modular application according to the enterprise design standards and accumulated design experience, and the special requirements and functional requirements are designed with high precision. All data calculations and data analysis are carried out manually. Therefore, there are large design errors, resulting in repeated drawing modifications for some positions or quality defect points of precision molds or complex molds, which not only delays the mold production time but also increases the mold production cost. Summary of the Invention

[0003] In order to solve the deficiencies in the prior art, the purpose of the present invention is to provide a method and device for rapid design of injection molds based on AI. By setting up various functional modules such as an information acquisition module, a model construction module, a model training module, a model evaluation module, a parameter determination module, and a determination scheme module, the quality abnormality information of daily similar injection molds in the industry is collected for system calculation and system analysis. To prevent quality problems such as "short shot, air entrapment, embrittlement, burning, flash, shrinkage, weld line, silver streak, warpage deformation, and dimensional deviation" in the development of injection molds from the source of design, preset the molding process parameters, perform automated production and detection, reduce human interference, and the core technical problems are completed by the system AI. Designers complete some auxiliary work such as marking numbers, which improves the mold design efficiency and reduces the mold production cost.

[0004] To achieve the above purpose, the technical solution adopted by the present invention is:

[0005] A method for rapid design of injection molds based on AI is applied to an injection mold design control system based on AI. The system includes an information acquisition module, a model construction module, a model training module, a model evaluation module, a parameter determination module, a determination scheme module, a wireless communication module, a memory, an alarm, a processing center, and an intelligent mobile terminal; the information acquisition module, the model construction module, the model training module, the model evaluation module, the parameter determination module, the determination scheme module, the wireless communication module, the memory, and the alarm are respectively connected to the processing center; the intelligent mobile terminal includes a smart phone, a tablet computer, and a smart remote control, and is wirelessly connected to the wireless communication module within the range of a wireless network or the Internet;

[0006] The wireless communication module is provided with a wireless network unit, which is responsible for the transceiver of wireless signals and automatically forms a network connection with the intelligent mobile terminal within an effective network range;

[0007] The alarm compares the actual evaluation score of the model with the model evaluation score standard stored in the memory. If it does not meet the standard, it will automatically give a sound alarm and notify to continue training; and compares the mold forming quality information with the injection mold quality standard stored in the memory. If it does not meet the standard, it will automatically give a sound alarm and notify to adjust the parameters for rework; and compares the quality information of the mold with the injection mold quality standard stored in the memory. If it does not meet the standard, it will automatically give a sound alarm and notify to rework again;

[0008] The memory is responsible for storing the information of the information acquisition module, model construction module, model training module, model evaluation module, parameter determination module, determination scheme module, wireless communication module, and alarm, as well as storing the model evaluation score standard and the injection mold quality standard;

[0009] The processing center is responsible for the information transfer of the information acquisition module, model construction module, model training module, model evaluation module, parameter determination module, determination scheme module, wireless communication module, alarm, and memory. It is the hub center of the system and compares the actual evaluation score of the model with the model evaluation score standard stored in the memory: if it meets the standard, it is the predetermined injection mold process parameters; if it does not meet the standard, it is transferred to the alarm and notifies to continue training; and compares the mold forming quality information with the injection mold quality standard stored in the memory: if it meets the standard, it is set as the formal injection mold process parameters; if it does not meet the standard, it is transferred to the alarm and notifies to adjust the parameters for rework; and compares the quality information of the mold with the injection mold quality standard stored in the memory: if it meets the standard, it notifies the customer that mass production can be carried out; if it does not meet the standard, it is transferred to the alarm and notifies to rework again;

[0010] The information acquisition module includes a data acquisition unit, a data cleaning unit, a data enhancement unit, an integration data unit, a data conversion unit, a filtering and denoising unit, a grayscale conversion unit, and a feature extraction unit. It is responsible for acquiring the historical image data of normal and abnormal mold forming qualities similar to the injection mold and performing preprocessing, and then transferring it to the model construction module;

[0011] The model construction module includes a network layer unit, a network parameter unit, a layer optimization unit, and a data division unit. It is responsible for determining the model architecture, designing the network layer structure, adding auxiliary layers, and defining model parameters according to the extracted feature information to construct the model, and then transferring it to the model training module;

[0012] The model training module includes an input filling unit, a convolution input unit, a pooling conversion unit, a fully connected layer unit, a normalization output unit, an output conversion unit, and a backpropagation unit, which are responsible for training the model according to the collected data, adjusting the model parameters to optimize the model performance for predetermined injection mold process parameters, and transmitting them to the model evaluation module;

[0013] The model evaluation module includes a classification marking unit, a learning and training unit, and a model evaluation unit, which are responsible for validating and evaluating the model according to the trained model, adjusting and optimizing the model according to the verification results, and transmitting them to the parameter determination module;

[0014] The parameter determination module includes a mold size unit, a forming pressure unit, an injection weight unit, an ejector pin strength unit, a mold manufacturing unit, a forming debugging unit, and an anomaly identification unit, which are responsible for conducting mold development tests according to the forming process parameters predicted by the trained model to obtain the optimal injection mold process parameters, and transmitting them to the determination scheme module;

[0015] The determination scheme module includes a shrinkage confirmation unit, a drawing file determination unit, and an appearance inspection unit, which are responsible for confirming the shrinkage rate according to the quality information of qualified injection molded parts, improving the mold design drawing file, and inspecting the appearance quality of the injection mold to ensure that the injection mold design meets the customer quality requirements.

[0016] A rapid design method for injection molds based on AI provided by the present invention includes the following steps:

[0017] S10. Before debugging, the information acquisition module acquires historical image data of normal and abnormal forming qualities of similar injection molds and performs preprocessing for subsequent model construction, and transmits them to the model construction module;

[0018] S20. The model construction module determines the model architecture, designs the network hierarchical structure, adds auxiliary layers, and defines model parameters according to the extracted mold feature information to construct the model for subsequent model training and verification, and transmits them to the model training module;

[0019] S30. The model training module trains the model according to the collected data, adjusts the model parameters to optimize the model performance for predetermined injection mold process parameters, ensures the quality stability of the injection mold design, and transmits them to the model evaluation module;

[0020] S40. The model evaluation module validates and evaluates the model according to the trained model, and adjusts and optimizes the model according to the verification results to improve the performance and accuracy of the model, and transmits them to the parameter determination module;

[0021] S50. The parameter determination module conducts trial mold verification based on the molding process parameters predicted by the trained model to obtain the optimal injection mold process parameters, ensuring the quality of subsequent molding trial production, and transmits them to the determination scheme module.

[0022] S60. The determination scheme module confirms the shrinkage rate based on the quality information of the injection molded parts that pass the trial mold verification, improves the mold design drawing, and inspects the appearance quality of the injection mold to ensure that the injection mold design meets the customer's quality requirements.

[0023] Furthermore, the step S10 includes the following steps:

[0024] S11. The data acquisition unit obtains historical data of various quality data, mold data, and their molding process parameters of similar injection moldings by connecting to the industry database, preprocesses them for subsequent data cleaning, and transmits them to the data cleaning unit.

[0025] S12. The data cleaning unit cleans the collected mold historical data to remove outliers, duplicate values, or missing values in the data, improving the quality and accuracy of the data, and transmits them to the data enhancement unit.

[0026] S13. The data enhancement unit increases the quantity and diversity of injection mold training data through data enhancement methods such as rotation, scaling, flipping, cropping, and color transformation to improve the model generalization ability, and transmits them to the integrated data unit.

[0027] S14. The integrated data unit obtains standardized injection mold data according to the data standardization processing formula "z = (x - μ) / σ, where z is the standardized data, x is the original data, μ is the mean of the data, and σ is the standard deviation of the data" to improve the stability of the model, and transmits them to the conversion data unit.

[0028] S15. The conversion data unit obtains normalized injection mold data according to the normalization calculation formula "x' = (x - min(x)) / (max(x) - min(x)), where x' is the normalized data, x is the original data, and min(x) and max(x) are the minimum and maximum values of the data respectively" to improve the model performance, and transmits them to the filtering and denoising unit.

[0029] S16. The filtering and denoising unit obtains the denoised injection mold design image according to the Gaussian filtering method calculation formula " G(x, y) is the pixel of the two-dimensional Gaussian function, (x, y) is the pixel coordinate, and σ is the standard deviation" for grayscale image conversion, and transmits them to the grayscale conversion unit.

[0030] S17. The grayscale conversion unit obtains the grayscale-injected mold design image according to the grayscale calculation formula "f(I, j) = max(R(I, j), G(I, j), B(I, j)), where f(I, j) is the image after grayscale, and R(I, j), G(I, j), and B(I, j) are the original images of the three colors" for subsequent feature extraction and transmits it to the feature extraction unit;

[0031] S18. The feature extraction unit converts the labeled data into feature vectors useful for model training and extracts the key features of the geometry, surface quality, dimensional accuracy, and draft angle of the mold and injection molded parts to improve the training speed and performance of the model.

[0032] Further, the step S20 includes the following steps:

[0033] S21. The network layer unit customizes and optimizes the network layer structure according to the selected convolutional neural network as the model architecture and based on the specific application scenario and requirements for subsequent defect detection and quality recognition, and transmits it to the network parameter unit;

[0034] S22. The network parameter unit adjusts the number of network layers, the size of the convolutional kernel, the stride, the padding method, and selects the loss function and optimization algorithm according to the network layer structure to ensure the performance and accuracy of the model, and transmits it to the hierarchical optimization unit;

[0035] S23. The hierarchical optimization unit adjusts the size of the input data and performs normalization processing by adding zero-padding layers and normalization layers before and after the convolutional layer respectively to improve the model training speed and stability, and transmits it to the data partitioning unit;

[0036] S24. The data partitioning unit divides the dataset with key feature vectors after extraction into a training set, a validation set, and a test set and establishes a convolutional neural network structure for subsequent model training and learning.

[0037] Further, the step S30 includes the following steps:

[0038] S31. The input padding unit obtains the padding data according to the mold data padding size calculation formula "Ph = [(Ho - 1) * Sh + Kh - Hi] / 2, Pw = [(Wo - 1) * Sw + Kw - Wi] / 2, where Ph and Pw are the padding sizes in the height / width directions, Hi and Wi are the height / width of the input feature map, Kh and Kw are the height / width of the convolutional kernel, Sh and Sw are the values of the stride in the height / width directions, and Ho and Wo are the height and width of the output feature map", inputs it, and transmits it to the convolutional input unit;

[0039] S32. The convolution input unit obtains the convolution input value according to the convolution input calculation formula "z(t) = ∫x(m)y(t - m)dm, where z(t) is the output function, x(t) is the input function, y(t) is the convolution kernel function, and dm is the integral element of the variable m", activates the function input, and transmits it to the pooling conversion unit;

[0040] S33. The pooling conversion unit obtains the maximum pixel value after pooling according to the max pooling calculation formula

[0041] "Z(I,j) = max(X[i*Ps(i + 1)*Ps,j*Ps(j + 1)*Ps])

[0042] , where Z(I,j) is a pixel value in the output feature map after pooling, i and j are the positions in the output feature map respectively, max is to obtain the maximum value among all pixel values in the input window, X is the input feature map, and Ps is the window size of the pooling operation", and enters the fully connected layer to retain important feature information, and transmits it to the fully connected layer unit;

[0043] S34. The fully connected layer unit obtains the output data of the connection layer according to the fully connected layer calculation formula "y = f(∑(Wn*Xn)+b), where y is the output result, f is the activation function, Wn is the weight of the nth input feature, Xn is the n input features, n is the dimension of the input features, and b is the bias", and enters the output layer for subsequent output calculation, and transmits it to the normalized output unit;

[0044] S35. The normalized output unit obtains the output of Batch Norm according to the Batch Norm layer calculation formula "h = φ(Bn(Wx + b)), where h is the output of the fully connected layer, φ is the activation function, Bn is the operator of batch normalization, W is the weight parameter, X is the input of the fully connected layer, and b is the bias parameter", and transmits it to the output conversion unit;

[0045] S36. The output conversion unit obtains the size of the output value after convolution according to the output layer conversion calculation formula "N = (P - F + 2C) / S + 1, where N is the output size after convolution, P is the input size before convolution, F is the convolution kernel size, C is the number of layers of adding 0 around the image, and S is the step size", for subsequent model analysis and training, and transmits it to the backpropagation unit;

[0046] S37. The backpropagation unit obtains the loss function value according to the loss function calculation formula "L = max(0, m + y*(f(x) - b)), where L is the loss function value, m is the hyperparameter, y is the sample label (0 or 1), f(x) is the predicted value of the model, and b is the classification boundary", and performs backpropagation to optimize the performance of the network model.

[0047] Further, the step S40 includes the following steps:

[0048] S41. The classification and marking unit divides and marks the corresponding data sample categories according to the quality abnormality categories of the injection molds for subsequent machine learning of the model and transmits them to the learning and training unit;

[0049] S42. The learning and training unit determines the training model according to the gradient descent data, inputs various mold quality data in the training set into the model for simulation training to ensure the accuracy of model recognition, and transmits them to the model evaluation unit;

[0050] S43. The model evaluation unit obtains the comprehensive evaluation score F according to the model evaluation score calculation formula "F = 2(Ac * Re) / (Ac + Re), where F is the evaluation score of the training model, Ac is the precision rate, and Re is the recall rate", and transmits it to the processing center;

[0051] S44. The processing center compares the actual evaluation score of the model with the model evaluation score standard stored in the memory: if it meets the standard, it is the predetermined injection mold process parameters; if it does not meet the standard, it is transmitted to the alarm and notifies to continue training until it meets the standard.

[0052] Further, the step S50 includes the following steps:

[0053] S51. The mold size unit obtains the injection mold size data according to the mold size calculation formulas "Dm 1 = (D M + Im + Es)*(1 + Za), Dm 2 = D M + Im + Es + Bh + Cg, Dm 3 = D M3 *(1 + So) - Cv, Dm 1 、Dm 2 、Dm 3 are the outer shape, structure, and cavity radial dimensions of the mold respectively, D M is the injection product size, D M3 is the radial limit size of the injection part, Im is the inner margin of the mold, Es is the shrinkage at both ends of the injection part, Za is the machining allowance of the mold, Bh is the height of the mold boss, Cg is the mold clamping gap, So is the shrinkage rate of the plastic material, and Cv is the mold correction value" for subsequent injection mold design marking and transmits them to the forming pressure;

[0054] S52. The forming pressure unit calculates the injection molding machine forming pressure according to the formula "Pi = P * Sa / [π*(d / 2) 2 , P F = Sm * Pv / 1000, where Pi is the injection pressure of the injection molding machine Pi (kg / cm 2 ), P is the pump pressure of the injection molding machine (kg / cm2 ), Sa is the effective area of the injection cylinder of the injection molding machine (cm 2 ), d is the diameter of the screw of the injection molding machine (cm), P F is the clamping force of the injection molding machine (T), Sm is the projected area of the mold cavity of the injection molding machine (cm 2 ), Pv is the filling pressure of the injection molding machine (kg / cm 2 )” respectively obtain the forming injection pressure and clamping pressure for subsequent mold forming debugging and transfer them to the injection weight unit;

[0055] S53. The injection weight unit obtains the injection weight of the injection molding machine formed according to the injection molding machine forming injection weight calculation formula "Wv = (π * (d / 2) 2 * Lt) * η * δ, where Wv is the injection weight of the injection molding machine (g), d is the diameter of the screw of the injection molding machine (cm), Lt is the injection stroke of the injection molding machine (cm), η is the specific gravity of the plastic material (g / cm 3 ), δ is the efficiency of the injection molding machine" to facilitate subsequent injection mold forming debugging and transfer it to the ejector pin strength unit;

[0056] S54. The ejector pin strength unit obtains the ejector pin strength of the injection mold according to the injection mold ejector pin strength calculation formula "F = π 2 * λ * (π * (d o / 2) 2 ) * E * r 2 / L 2 , where F is the buckling load of the mold ejector pin (kgf), λ is the mold ejector pin support condition constant, d o is the diameter of the mold ejector pin (cm), E is the longitudinal elastic modulus of the mold ejector pin (kgf / cm 2 ), r is the sectional radius of inertia of the mold ejector pin (cm), L is the length of the mold ejector pin (cm)" to facilitate subsequent injection mold forming debugging and transfer it to the mold manufacturing unit;

[0057] S55. The mold manufacturing unit designs drawings according to the predetermined mold manufacturing process parameters, manufactures the mold through mold manufacturing equipment, and obtains mold processing quality information to ensure the quality of subsequent forming trials and transfers it to the forming debugging unit;

[0058] S56. The forming debugging unit conducts forming debugging through the injection molding machine according to the predetermined forming process parameters and obtains the quality information of the injection molded parts to ensure the accuracy of the mold design and transfers it to the processing center;

[0059] S57. The processing center compares the mold quality information with the quality standard of this injection mold stored in the memory: if it meets the standard, it is set as the formal injection mold process parameters; if it does not meet the standard, it is transferred to the alarm and the parameter adjustment and rework are notified.

[0060] S58. The anomaly recognition unit matches the quality inspection image of the injection mold obtained in real time with the corresponding image in the trained convolutional neural network model and confirms its anomaly category for subsequent molding improvement and parameter adjustment.

[0061] Further, step S60 includes the following steps:

[0062] S61. The shrinkage confirmation unit obtains the shrinkage rate of the injection molded part according to the shrinkage rate calculation formula of the injection molded part "Sr = [-D M +sqrt(4D M *Dm - 3D M 2 )] / 2D M , where Sr is the shrinkage rate, Dm is the mold size, and D M is the size of the injection molded part" to confirm the accuracy of the injection molded part and transfer it to the appearance inspection unit;

[0063] S62. The drawing determination unit improves the design drawing of the injection mold and its corresponding design information according to the parameters after trial mold debugging to determine the final mold design plan and transfer it to the appearance inspection unit;

[0064] S63. The appearance inspection unit obtains the image or video of the injection mold through a high-definition camera and converts it into recognizable appearance quality information for subsequent packaging and shipping and mass production of molding, and transfers it to the processing center;

[0065] S64. The processing center compares the quality information of the mold with the quality standard of this injection mold stored in the memory: if it meets the standard, it notifies the customer that mass production can be carried out; if it does not meet the standard, it transfers it to the alarm and notifies rework.

[0066] The system of the present invention further includes a computer-readable storage medium containing a memory; the memory stores a computer program, and when each functional module executes the computer program, it realizes the steps of any one of the above-mentioned AI-based rapid injection mold design methods; a computer program is stored on the computer-readable storage medium, and when the computer program is executed by each functional module, it realizes the steps of any one of the above-mentioned AI-based rapid injection mold design methods.

[0067] The present invention also provides an AI-based injection mold design control device implemented by using any one of the above-mentioned AI-based rapid injection mold design methods.

[0068] The beneficial effects of the present invention compared with the prior art:

[0069] By setting up each functional module, collecting the abnormal quality information of daily similar injection molds in the industry for system calculation and analysis, preventing quality problems such as "short shot, gas entrapment, brittleness, burning, flash, shrinkage, weld line, silver streak, warping deformation, and dimensional deviation" from the design source during the development of injection molds, presetting the forming process parameters, and performing automated mold manufacturing and detection to reduce human interference, the core technical problems are completed by the system AI, and the designers complete some auxiliary work such as marking numbers, thus improving the mold design efficiency and reducing the mold manufacturing cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or exemplary technical descriptions. The following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0071] Figure 1 Schematic diagram of the system module of the present invention;

[0072] Figure 2 Schematic diagram of the information acquisition module of the present invention;

[0073] Figure 3 Schematic diagram of the model construction module of the present invention;

[0074] Figure 4 Schematic diagram of the model training module of the present invention;

[0075] Figure 5 Schematic diagram of the model evaluation module of the present invention;

[0076] Figure 6 Schematic diagram of the parameter determination module of the present invention;

[0077] Figure 7 Schematic diagram of the determination scheme module of the present invention;

[0078] Figure 8 Schematic diagram of the method flow control program of the present invention;

[0079] Figure 9 Schematic diagram of the program of step S10 in the method flow of the present invention;

[0080] Figure 10 Schematic diagram of the program of step S20 in the method flow of the present invention;

[0081] Figure 11 Schematic diagram of the program of step S30 in the method flow of the present invention;

[0082] Figure 12It is a schematic diagram of the procedure of step S40 in the method flow of the present invention;

[0083] Figure 13 It is a schematic diagram of the procedure of step S50 in the method flow of the present invention;

[0084] Figure 14 It is a schematic diagram of the procedure of step S60 in the method flow of the present invention; Detailed implementation manners

[0085] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0086] The following describes the specific implementation of the present invention in detail with specific embodiments:

[0087] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. It should be noted that when a module is referred to as being "disposed on" another module, it can be directly on the other module or indirectly on the other module. When a module is referred to as being "connected to" another module, it can be directly connected to the other module or indirectly connected to the other module.

[0088] In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically defined. The meaning of "several" is one or more, unless otherwise specifically defined. The "injection mold" and "mold" in the present application both refer to injection molding molds, and the "injection molded part", "product", and "injection molded product" all refer to injection molded parts formed by an injection mold.

[0089] In the description of the present application, it should be noted that unless otherwise clearly specified and limited, the terms "connected" and "connection" 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 directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two components or the interaction relationship between two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific situations. The terms "including", "comprising", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways; the "data" in the present application all refer to the data of an injection mold.

[0090] Please refer to Figure 1As shown in the figure, the present invention provides a rapid design method for injection molds based on AI, which is applied to an AI-based injection mold design control system. The system includes an information acquisition module, a model construction module, a model training module, a model evaluation module, a parameter determination module, a determination scheme module, a wireless communication module, a memory, an alarm, a processing center, and an intelligent mobile terminal. The information acquisition module, the model construction module, the model training module, the model evaluation module, the parameter determination module, the determination scheme module, the wireless communication module, the memory, and the alarm are respectively connected to the processing center. The intelligent mobile terminal includes a smart phone, a tablet computer, and a smart remote control, and is wirelessly network-connected to the wireless communication module within the range of a wireless network or the Internet.

[0091] The wireless communication module is provided with a wireless network unit, which is responsible for the transceiver of wireless signals and automatically forms a network connection with the intelligent mobile terminal within an effective network range. The wireless signals include various Internet of Things signals such as MQTT, CoAP, HTTP, REST API, Zagbee, LoRaWAN, NB-IoT, Bluetooth, etc., or 5G, 4G, Wifi network signals.

[0092] The alarm compares the actual model evaluation score with the model evaluation score standard stored in the memory. If the standard is not met, it will automatically give an audible alarm and notify to continue training. It also compares the mold forming quality information with the injection mold quality standard stored in the memory. If the standard is not met, it will automatically give an audible alarm and notify to adjust the parameters and rework. And it compares the quality information of the mold with the injection mold quality standard stored in the memory. If the standard is not met, it will automatically give an audible alarm and notify to rework again.

[0093] The memory is responsible for storing the information of the information acquisition module, the model construction module, the model training module, the model evaluation module, the parameter determination module, the determination scheme module, the wireless communication module, and the alarm, as well as storing the model evaluation score standard and the injection mold quality standard.

[0094] The processing center is responsible for the information transfer of the information acquisition module, model construction module, model training module, model evaluation module, parameter determination module, determination scheme module, wireless communication module, alarm, and memory. It is the hub center of the system and compares the actual evaluation score of the model with the model evaluation score standard stored in the memory: if it meets the standard, it is the predetermined injection mold process parameters; if it does not meet the standard, it is transmitted to the alarm and notifies to continue training. It also compares the mold forming quality information with the injection mold quality standard stored in the memory: if it meets the standard, it is set as the formal injection mold process parameters; if it does not meet the standard, it is transmitted to the alarm and notifies to adjust the parameters and rework. It further compares the quality information of the mold with the injection mold quality standard stored in the memory: if it meets the standard, it notifies the customer that mass production can be carried out; if it does not meet the standard, it is transmitted to the alarm and notifies to rework again.

[0095] Please refer to Figure 2 As shown, the information acquisition module includes a data acquisition unit, a data cleaning unit, a data enhancement unit, an integration data unit, a conversion data unit, a filtering and denoising unit, a grayscale conversion unit, and a feature extraction unit. It is responsible for acquiring the normal and abnormal historical image data of the forming quality of similar injection molds and performing preprocessing, and then transmitting it to the model construction module.

[0096] Furthermore, the data acquisition unit obtains the historical data of various quality data, mold data, and their forming process parameters of similar injection mold forming by connecting to the industry database and performs preprocessing, and then transmits it to the data cleaning unit; the data cleaning unit cleans the collected mold historical data to remove outliers, duplicate values, or missing values in the data to improve the quality and accuracy of the data, and then transmits it to the data enhancement unit; the data enhancement unit increases the quantity and diversity of injection mold training data through data enhancement methods such as rotation, scaling, flipping, cropping, and color transformation, and then transmits it to the integration data unit; the integration data unit obtains the standardized injection mold data according to the data standardization processing formula "z=(x - μ) / σ", and then transmits it to the conversion data unit; the conversion data unit obtains the normalized injection mold data according to the normalization calculation formula "x'=(x - min(x)) / (max(x) - min(x))", and then transmits it to the filtering and denoising unit; the filtering and denoising unit obtains the denoised injection mold design image and transmits it to the grayscale conversion unit; the grayscale conversion unit obtains the grayscaled injection mold design image according to the grayscaling calculation formula "f(I,j)=max(R(I,j),G(I,j),B(I,j))", and then transmits it to the feature extraction unit; the feature extraction unit converts the labeled data into feature vectors useful for model training and extracts the key features of the geometry, surface quality, dimensional accuracy, and draft angle of the mold and injection parts to improve the training speed and performance of the model.

[0097] Please refer to Figure 3 As shown, the model construction module includes a network layer unit, a network parameter unit, a hierarchical optimization unit, and a data division unit, which are responsible for constructing a model by determining the model architecture, designing the network layer structure, adding auxiliary layers, and defining model parameters according to the extracted feature information, and transmitting it to the model training module.

[0098] Furthermore, the network layer unit customizes and optimizes the network layer structure according to the selected convolutional neural network as the model architecture and based on specific application scenarios and requirements, and transmits it to the network parameter unit; the network parameter unit adjusts the number of network layers, the size of the convolutional kernel, the stride, the padding method, and selects the loss function and optimization algorithm according to the network layer structure, and transmits it to the hierarchical optimization unit; the hierarchical optimization unit adjusts the size of the input data and performs normalization processing by adding zero-padding layers and normalization layers before and after the convolutional layer respectively, and transmits it to the data division unit; the data division unit divides the dataset with key feature vectors after extraction into a training set, a validation set, and a test set and establishes a convolutional neural network structure for subsequent model training and learning.

[0099] Please refer to Figure 4 As shown, the model training module includes an input padding unit, a convolutional input unit, a pooling conversion unit, a fully connected layer unit, a normalized output unit, an output conversion unit, and a backpropagation unit, which are responsible for training the model according to the collected data, adjusting the model parameters, and optimizing the model performance to the predetermined injection mold process parameters, and transmitting it to the model evaluation module.

[0100] Furthermore, the input filling unit obtains filling data according to the mold data filling size calculation formula "Ph = [(Ho - 1) * Sh + Kh - Hi] / 2, Pw = [(Wo - 1) * Sw + Kw - Wi] / 2", inputs it, and transfers it to the convolution input unit; the convolution input unit obtains the convolution input value according to the convolution input calculation formula "z(t) = ∫x(m)y(t - m)dm", activates the function input, and transfers it to the pooling conversion unit; the pooling conversion unit obtains the maximum pixel value after pooling according to the maximum pooling calculation formula "Z(I,j) = max(X[i*Ps(i + 1)*Ps,j*Ps(j + 1)*Ps])" and enters the fully connected layer, and transfers it to the fully connected layer unit; the fully connected layer unit obtains the output data of the connection layer according to the fully connected layer calculation formula "y = f(∑(Wn*Xn) + b)", enters the output layer, and transfers it to the normalization output unit; the normalization output unit obtains the output of Batch Norm according to the Batch Norm layer calculation formula "h = φ(Bn(Wx + b))", and transfers it to the output conversion unit; the output conversion unit obtains the size of the output value after convolution according to the output layer conversion calculation formula "N = (P - F + 2C) / S + 1", and transfers it to the backpropagation unit; the backpropagation unit obtains the loss function value according to the loss function calculation formula "L = max(0, m + y*(f(x) - b))" and performs backpropagation to optimize the performance of the network model.

[0101] Please refer to Figure 5 As shown, the model evaluation module includes a classification marking unit, a learning and training unit, and a model evaluation unit, which are responsible for validating and evaluating the model according to the trained model, adjusting and optimizing the model according to the verification results, and transferring it to the parameter determination module.

[0102] Furthermore, the classification marking unit divides the corresponding data sample categories according to the abnormal categories of the injection mold quality and marks them, and transfers them to the learning and training unit; the learning and training unit determines the training model according to the gradient descent data, inputs various mold quality data in the training set into the model for simulation training, and transfers it to the model evaluation unit; the model evaluation unit obtains the comprehensive F evaluation score according to the model evaluation score calculation formula "F = 2(Ac*Re) / (Ac + Re)", and transfers it to the processing center.

[0103] Please refer to Figure 6 As shown, the parameter determination module includes a mold size unit, a molding pressure unit, an injection weight unit, an ejector pin strength unit, a mold manufacturing unit, a molding debugging unit, and an abnormality identification unit, which are responsible for obtaining the best injection mold process parameters through mold development tests according to the molding process parameters predicted by the trained model, and transferring them to the determination scheme module.

[0104] Further, the mold size unit obtains injection mold size data according to the mold size calculation formula "Dm 1 =(D M +Im+Es)*(1+Za), Dm 2 =D M +Im+Es+Bh+Cg, Dm 3 =D M3 *(1+So)-Cv" and transmits it to the forming pressure; the forming pressure unit obtains the forming injection pressure and the clamping pressure respectively according to the injection molding machine forming pressure calculation formula "Pi = P*Sa / [π*(d / 2) 2 , P F =Sm*Pv / 1000" and transmits them to the injection weight unit; the injection weight unit obtains the injection weight of the injection molding machine forming according to the injection molding machine forming injection weight calculation formula "Wv=(π*(d / 2) 2 *Lt)*η*δ" and transmits it to the ejector pin strength unit; the ejector pin strength unit obtains the ejector pin strength of the injection mold according to the injection mold ejector pin strength calculation formula "F = π 2 *λ*(π*(d o / 2) 2 )*E*r 2 / L 2 " and transmits it to the mold manufacturing unit; the mold manufacturing unit designs drawings according to the predetermined mold manufacturing process parameters, manufactures the mold through the mold manufacturing equipment and obtains the mold processing quality information, and transmits it to the forming debugging unit; the forming debugging unit conducts forming debugging through the injection molding machine according to the predetermined forming process parameters and obtains the quality information of the injection molded part, and transmits it to the processing center; the abnormality recognition unit matches the real-time obtained injection mold quality inspection image with the corresponding image in the trained convolutional neural network model and confirms its abnormality category for subsequent forming improvement and parameter adjustment.

[0105] Please refer to Figure 7 shown, the determination scheme module includes a shrinkage confirmation unit, a drawing file determination unit, and an appearance detection unit, which are responsible for confirming the shrinkage rate according to the quality information of the injection molded part qualified by the trial mold verification, improving the mold design drawing file, and detecting the appearance quality of the injection mold to ensure that the injection mold design meets the customer quality requirements.

[0106] Further, the shrinkage confirmation unit calculates the shrinkage rate of the injection molded part according to the formula "Sr = [-D M +sqrt(4D M *Dm - 3D M 2 )] / 2D M”Obtain the shrinkage rate of the injection molded part and transmit it to the appearance inspection unit; the drawing determination unit improves the injection mold design drawing and its corresponding design information according to the parameters after trial mold debugging to determine the final mold design plan, and transmits it to the appearance inspection unit; the appearance inspection unit obtains the image or video of the injection mold through a high-definition camera and converts it into recognizable appearance quality information for subsequent packaging and shipping and mass production, and transmits it to the processing center.

[0107] System operation principle:

[0108] Before debugging, the information acquisition module acquires the historical image data of normal and abnormal forming quality of similar injection molds, preprocesses it, and transmits it to the model construction module; then the model construction module determines the model architecture, designs the network hierarchy structure, adds auxiliary layers, and defines model parameters according to the extracted mold feature information to construct the model, and transmits it to the model training module; the model training module trains the model according to the collected data, adjusts the model parameters to optimize the model performance with predetermined injection mold process parameters, and transmits it to the model evaluation module; then the model evaluation module verifies and evaluates the model according to the trained model, adjusts and optimizes the model according to the verification results, and transmits it to the parameter determination module; the parameter determination module obtains the best injection mold process parameters through trial mold verification according to the forming process parameters predicted by the trained model, and transmits it to the determination plan module; then the determination plan module confirms the shrinkage rate according to the quality information of the injection molded part qualified by trial mold verification, improves the mold design drawing, and inspects the appearance quality of the injection mold to ensure that the injection mold design meets the customer quality requirements.

[0109] When designers or managers are outdoors or in other places, they can use intelligent mobile terminals to automatically form a network connection with the wireless communication module within the wireless network or the Internet, realizing the combination of intelligent mobile terminals, the Internet and Internet of Things technologies. Through the APP software on the intelligent mobile terminal or the remote control software on the computer, they can control the mold trial production situation at close range or remotely to meet the user quality requirements. Designers or managers can remotely control and monitor the mold trial verification situation of the injection mold, view the daily operation status, data change situation, equipment activity log, etc. By means of prefabricated deployment codes or adding deployment codes, remote connection and control of the injection mold trial verification can be realized. The trial mold results of the remote injection mold can be viewed in real time through both intelligent mobile terminals and computers, improving the trial mold efficiency and reducing the mold production cost.

[0110] Please refer to Figure 8 As shown, a rapid design method for injection molds based on AI provided by the present invention includes the following steps:

[0111] S10. Before debugging, the information acquisition module acquires historical image data of normal and abnormal molding quality of similar injection molds, preprocesses it for subsequent model construction, and transfers it to the model construction module;

[0112] Please refer to Figure 9 As shown, the step S10 includes the following steps:

[0113] S11. The data acquisition unit acquires historical data of various quality data, mold data, and their molding process parameters of similar injection moldings by connecting to the industry database, preprocesses it for subsequent data cleaning, and transfers it to the data cleaning unit;

[0114] Further explanation: According to the customer order information, collect historical data of similar injection molding process parameters from domestic and foreign injection mold databases through a data collector, as well as the quality information of images or videos of the corresponding injection molds and their injection molded parts, and classify and aggregate the data of the trial mold verification effects of different users; The injection mold quality data includes, but is not limited to, historical data of the trial mold verification effect evaluations of different users for different such injection molds, and also includes normal and abnormal quality information of similar injection moldings obtained from competitors through legal means such as web crawling, sensor collection, manual annotation, dataset purchase, and crowdsourcing from third-party shared data platforms such as the Internet, social media, professional testing institutions, injection mold trading platforms, and public databases, as well as the historical data of the trial mold verification effects and injection molded part assembly application effects of different users, in order to obtain high-quality, diverse, and rich historical data of similar injection mold quality, specifically including the geometric dimensions, structural features, material types, molding process parameters, and historical injection mold production data of the mold and its injection molded parts.

[0115] S12. The data cleaning unit cleans the collected mold historical data to remove outliers, duplicate values, or missing values in the data, to improve the quality and accuracy of the mold data, and transfers it to the data enhancement unit;

[0116] Further explanation: Outliers may occur due to sensor failures, recording errors, or system anomalies during the injection mold forming verification, which may affect the accuracy of subsequent analysis. Missing value processing is performed on data with missing parts. Code is written to detect each feature in the dataset to determine missing values and identify missing value patterns. According to the number and impact of missing values, code is written to select whether to fill or delete samples or variables with missing values. For time series data, forward filling or backward filling is used to fill in the missing values. The processing effect is verified by comparing indicators such as data quality, model accuracy, and reliability before and after processing. If the processing effect is not good, the processing method is reselected or the data is cleaned again until the best effect is achieved to remove duplicate, incorrect, and invalid data to effectively process missing values for subsequent model construction and analysis. The injection mold quality data includes information such as the appearance, dimensions, structure, material and heat treatment, and injection performance of the injection mold, as well as information such as the appearance quality, dimensional accuracy, physical and chemical properties of the injection molded parts, and also includes operation parameters of the mold manufacturing equipment, injection molding process parameters, and mold manufacturing processes.

[0117] S13. The enhanced data unit uses data augmentation methods such as rotation, scaling, flipping, cropping, and color transformation to increase the quantity and diversity of injection mold training data, improve the generalization ability of the model, and transfer it to the integrated data unit;

[0118] Further explanation: After rotating the cleaned mold image data by a certain angle through mold image data rotation, multiple pixels may correspond to the same pixel after rotation, resulting in the loss of pixels in the rotated mold image data. Therefore, a reverse thinking is used to construct a mapping to map the pixel coordinates after rotation to the pixel coordinates of the original mold image data, which can ensure that each pixel after rotation has a pixel value. The neighborhood interpolation algorithm is used to copy each original pixel in the rotated injection mold data unchanged to the corresponding four pixels after expansion, retaining all the information of the original mold image data. The mold image data with abnormal orientation after scaling is flipped 180 degrees around the center axis or symmetry axis of the original image to ensure the orientation consistency of the injection mold data. The flipped mold image data is cropped to ensure the integrity and clarity of the image.

[0119] S14. The integrated data unit obtains the standardized injection mold data according to the data standardization processing formula "z=(x - μ) / σ, where z is the standardized data, x is the original data, μ is the mean of the data, and σ is the standard deviation of the data" to improve the stability of the model and transfer it to the conversion data unit;

[0120] Further explanation: Calibrate the original data of the injection mold collected, including time synchronization, unit conversion, range adjustment, etc. For example, ensure that all timestamps are in a unified format and synchronized with the system clock; convert the data collected by different sensors, detection devices, etc. to the same unit to ensure the accuracy of the data; use the above data standardization processing formula "z = (x - μ) / σ" through the written code to convert data from different sources such as dates, values, and texts into a standard normal distribution with a mean of 0 and a standard deviation of 1, subtract the mean and divide by the standard deviation to make it in a unified standard format to ensure the consistency and comparability of the data for subsequent processing; encode unstructured data to convert it into structured data, or convert text data into numerical representations for easy model processing and analysis, so as to process the original data into a form that conforms to a specific standard for subsequent model training, learning, and analysis.

[0121] S15. The data conversion unit obtains the normalized injection mold data according to the normalization calculation formula "x' = (x - min(x)) / (max(x) - min(x)), where x' is the normalized data, x is the original data, and min(x) and max(x) are the minimum and maximum values of the data respectively" to improve the model performance and transfer it to the filtering and denoising unit;

[0122] Further explanation: Through the written code, use the above normalization calculation formula "x' = (x - min(x)) / (max(x) - min(x))" to scale the cleaned data proportionally to the specified range of [0, 1], convert the data to a unified format or range for data normalization, and perform data discretization such as converting continuous features to discrete features to transform the data, so as to eliminate the total difference between different samples or features and the dimensional influence between data, make the distribution of the data consistent, and confirm whether the data has been normalized to the specified range or distribution as expected, and verify the normalization effect through statistical descriptions such as the maximum value and minimum value; the data includes various types such as text, pictures, audio, and video to eliminate the dimensional differences between different data, thereby improving the accuracy and efficiency of data analysis to better adapt to subsequent analysis and processing.

[0123] S16. The filtering and denoising unit obtains the denoised injection mold design image according to the Gaussian filtering method calculation formula " G(x, y) is the pixel of the two-dimensional Gaussian function, (x, y) is the pixel coordinate, and σ is the standard deviation" for grayscale image conversion and transfers it to the grayscale conversion unit;

[0124] Further explanation: The Gaussian filter output pixel value obtained according to the Gaussian filter calculation formula is the weighted average of the input pixel values within its neighborhood, and the weights are given by the Gaussian function. Among them, the standard deviation σ determines the width of the Gaussian function, and the Gaussian function is discretized, that is, the weights are calculated within a certain window size, and then these weights are applied to the corresponding neighborhood pixel values of the input image to obtain the output pixel value. The pixels are arranged in ascending order according to the gray value, and the median value is taken as the new gray value; check the sampled values in the input signal to determine whether they represent the signal itself. By using an observation window composed of an odd number of samples, the values within the window are sorted, the middle value is taken as the output, the earliest value is discarded, and a new sample is obtained. Repeat the above calculation process to reduce the random noise in the image and make the image smoother.

[0125] S17. The grayscale conversion unit obtains the grayscale injection mold design image according to the grayscale calculation formula "f(I, j) = max(R(I, j), G(I, j), B(I, j)), where f(I, j) is the image after grayscale conversion, and R(I, j), G(I, j), and B(I, j) are the original images of the three colors" for subsequent feature extraction and transfers it to the feature extraction unit;

[0126] Further explanation: The grayscale injection mold design image is obtained through the grayscale calculation formula. Check the sampled values in the input signal to determine whether they represent the signal itself. By using an observation window composed of an odd number of samples, the values within the window are sorted, the middle value is taken as the output, the earliest value is discarded, and a new sample is obtained. Repeat the above calculation process to remove the noise in the injection mold design image or other signals for subsequent extraction of the key features of the injection mold development image, including the geometric structures (including wall thickness, ribs, fillets, holes) of the injection mold and the injection molded part, accuracy, surface quality, dimensional accuracy, draft angle, etc. information; Then, the standardized data is classified through the written code, and the classified data is automatically labeled to add accurate labels or annotations to the data as the target variables or features during subsequent model training, and used as the training set and validation set to ensure that the model has accurate data reference during the training process, enabling it to learn how to generate corresponding labels according to the data features to improve the labeling efficiency and accuracy; For some data, the user needs to remotely add new labels or modify the existing labels through the mobile phone to enter the system to ensure that they accurately reflect the essential features of the data. The marked data is deployed to the corresponding system platform for subsequent application or analysis to improve the accuracy and reliability of the labels, facilitate the extraction of key features in the subsequent data, and be more conducive to model training, testing, and verification.

[0127] S18. The feature extraction unit converts the labeled data into feature vectors useful for model training and extracts the key features of the geometry, surface quality, dimensional accuracy, and draft angle of the mold and injection molded part, so as to improve the training speed and performance of the model.

[0128] Further explanation: Select the most representative features that can reflect the essential content of the data, have distinctiveness and descriptiveness, and are most useful for the model from the preprocessed injection mold data, reduce the dimension of the features, reduce the computational complexity, and at the same time improve the performance and generalization ability of the model; After feature extraction, principal component analysis is used for feature dimensionality reduction to obtain a large amount of feature information, in order to reduce the dimension of the features and reduce the computational complexity; After feature dimensionality reduction, the features are binary encoded to improve the interpretability of the features and the performance of the model; Since some information of the original data may be lost after feature extraction and dimensionality reduction, feature reconstruction such as principal component reconstruction or least squares reconstruction is performed in order to restore the information of the data as much as possible; Evaluation is carried out through cross-validation, and the feature extraction method and parameters are adjusted according to the evaluation results to evaluate the quality and effect of the extracted features, improve the training speed and performance of the model, and improve the efficiency of data analysis by reducing the quantity or complexity of the data while keeping the data as original as possible.

[0129] S20. The model construction module determines the model architecture, designs the network layer structure, adds auxiliary layers, and defines model parameters according to the extracted mold feature information to construct the model for subsequent model training and verification, and transmits it to the model training module;

[0130] Please refer to Figure 10 As shown, the step S20 includes the following steps:

[0131] S21. The network layer unit customizes and designs and optimizes the network layer structure according to the selected convolutional neural network as the model architecture and based on the specific application scenario and requirements for subsequent defect detection and quality identification, and transmits it to the network parameter unit;

[0132] Further explanation: Since the convolutional neural network (CNN) is a type of deep neural network, it is particularly suitable for processing data with a grid topology. In the development of injection molds, it can automatically learn and extract useful features such as edges, lines, corners, and more complex combined features from images for subsequent defect detection, quality classification, etc. Since the basic structure of the CNN consists of an input layer, convolutional layers, activation function layers, pooling layers, fully connected layers, and an output layer, these levels can work together in the CNN model for injection mold development to extract and process image features, including: The input layer can receive image data for injection mold development as input; multiple convolutional layers are set in the convolutional layer, and each layer contains multiple convolutional kernels, which can extract local features in the image. As the network structure deepens, it gradually becomes more abstract from the appearance; adding a ReLU activation function layer after the convolutional layer can introduce non-linear factors and enhance the expressive ability of the model; adding a pooling layer after the convolutional layer can reduce the dimension of the feature map, reduce the amount of calculation, and at the same time maintain the spatial invariance of important features; adding a fully connected layer at the end of the network is used to convert the feature maps output by the convolutional layer and the pooling layer into classification results, that is, the category or quality level of the injection mold design.

[0133] S22. The network parameter unit adjusts the number of network layers, the size of the convolutional kernel, the stride, the padding method, and selects the loss function and optimization algorithm according to the network hierarchical structure to ensure the performance and accuracy of the model, and transmits them to the hierarchical optimization unit;

[0134] Further explanation: Defining model parameters according to the network hierarchical design, including the size of the convolutional kernel, the stride, the padding method, the activation function, etc., as well as the pooling method of the pooling layer, the size of the pooling window, etc., and the number of neurons in the fully connected layer and other hierarchical parameters directly affect the performance and accuracy of the model; then set hyperparameters related to model training such as the size of the input image, the number of types of labels, the total number of training cycles, the batch size, etc., and adjust them according to the specific dataset and task requirements; select optimization algorithms such as SGD, Adam, etc., and set corresponding parameters such as the learning rate, momentum, etc., which directly affect the training effect and convergence speed of the model to ensure that the model can accurately extract features from the image and perform effective classification or prediction.

[0135] S23. The hierarchical optimization unit adjusts the size and normalization processing of the input data by adding zero-padding layers and normalization layers before and after the convolutional layer respectively to improve the training speed and stability of the model, and transmits them to the data partitioning unit;

[0136] Further explanation: By setting a zero-padding layer to perform zero-padding on the edges of the input image before the convolution operation, the size of the input data is adjusted to facilitate the convolution operation and preserve the image edge information, ensuring that the convolution kernel can be correctly applied to the image boundary. At the same time, it helps to preserve the image edge information and avoid information loss during the convolution process. Through zero-padding, input images of different sizes can be processed more effectively without complex preprocessing, which helps to control the size of the output feature map of the convolution layer, making the network design more flexible and controllable. It can ensure that the network can adapt to the design changes of injection molds with different sizes and resolutions and accurately extract and process image features. Normalize the distribution of the input data to improve the training speed and stability of the model, accelerate the training process of the neural network and improve the convergence speed. At the same time, enhance the stability of the model. Before the activation function of each layer of the network, normalize the input of the activation function. Calculate the mean and variance of this batch for each small batch of data, and then perform batch normalization on the linear calculation results, that is, subtract the mean and divide by the standard deviation to ensure that the calculation results conform to the standard normal distribution with a mean of 0 and a variance of 1. Then perform translation and scaling operations to adapt to different data distribution requirements, keeping the input of the middle layer of the network relatively stable, which helps to solve the problems of gradient disappearance or gradient explosion during the training process, thereby accelerating the training and enhancing the stability of the model. It can improve the robustness of the model to the weight initialization method, reduce the trouble caused by the weight initialization selection, and improve the generalization ability and training efficiency of the model.

[0137] S24. The data division unit divides the dataset with key feature vectors after extraction into a training set, a validation set, and a test set and establishes a convolutional neural network structure for subsequent model training and learning.

[0138] Further explanation: According to the preselected segmentation strategy, determine the proportions of the training set, validation set, and test set, and ensure that the data distributions among the subsets are as consistent as possible to avoid introducing biases. In some application scenarios, due to the time-series characteristics of the data, write code to divide the dataset into a training set for training the model, a validation set for adjusting model parameters and selecting the best model, and a test set for evaluating the final performance of the model according to the time series to ensure the temporal consistency of the training and test sets. Establish a training set and a test set required for predicting the abnormal recognition pattern of the injection mold design quality in different environments by passing the dataset through a convolutional neural network deep learning method with a 9:1 ratio, thereby establishing a deep convolutional neural network structure: 3 convolutional layers, 3 max pooling layers, 2 fully connected layers, 1 softmax layer, and 1 output layer. Add zero-padding layers and normalization layers when necessary. Since the convolutional kernels in the convolutional layers contain weight coefficients while the pooling layers do not, the pooling layers are not independent layers. The convolutional layers are used to extract local features of the image, the pooling layers are used to reduce the dimension of the feature maps, and the fully connected layers are used to integrate features and perform classification.

[0139] S30. The model training module trains the model based on the collected data, adjusts the model parameters to optimize the model performance for predetermined injection mold process parameters, ensures the quality stability of the injection mold design, and transfers it to the model evaluation module.

[0140] Please refer to Figure 11 As shown, the step S30 includes the following steps:

[0141] S31. The input padding unit obtains padding data according to the mold data padding size calculation formula "Ph = [(Ho - 1) * Sh + Kh - Hi] / 2, Pw = [(Wo - 1) * Sw + Kw - Wi] / 2, where Ph and Pw are the padding sizes in the height / width directions respectively, Hi and Wi are the height / width of the input feature map respectively, Kh and Kw are the height / width of the convolutional kernel respectively, Sh and Sw are the values of the stride in the height / width directions respectively, and Ho and Wo are the height and width of the output feature map respectively", inputs it, and transfers it to the convolutional input unit.

[0142] Further explanation: After normalizing the image, the injection mold data first enters the input layer, and the filling data is obtained through the above-mentioned filling size calculation formula and enters the zero-padding layer. Zero values are padded to the edges of the input data of the convolutional layer to adjust the size of the input data for convolutional operations, so as to maintain the same spatial dimension of the input and output, thereby preventing the loss of image edge information and enabling the spatial dimension of the input data to remain unchanged during subsequent convolutional operations, so as to preserve the image edge information and the processing of subsequent layers; the number of padding is jointly determined by the size of the convolutional kernel, the stride, and the padding size. In order to keep the height and width of the input and output images the same, or to reduce the rapid reduction of the size during consecutive convolutional operations, the number is usually set to the convolutional kernel size minus; the height Hi / width Wi of the input feature map, the height Kh / width Kw of the convolutional kernel, the stride values Sh / Sw in the height / width direction, and the height Ho / width Wo of the output feature map are all obtained through model design parameters.

[0143] S32. The convolutional input unit obtains the convolutional input value according to the convolutional input calculation formula "z(t) = ∫x(m)y(t - m)dm, where z(t) is the output function, x(t) is the input function, y(t) is the convolutional kernel function, and dm is the integration element for the variable m", activates the function input, and passes it to the pooling conversion unit;

[0144] Further explanation: The "width" or "increment" of the integration element dm when integrating each infinitesimal interval of the variable m is used to obtain the cumulative effect z(t) over the entire domain by accumulating the function values (i.e., x(m)y(t - m)) on these infinitesimal intervals. dm is the integration element for the variable m, that is, the infinitesimal quantity in the integration process; and the convolution calculation is performed to obtain the convolution value, which enters the convolutional layer and passes through the ReLU function calculation formula "f(x) = max(0, x + Y)" to obtain the activation function value and enter the activation function layer. In the formula, f(x) is the activated function, x is the eigenvalue output by the convolutional layer or the fully connected layer, and Y is a random variable. When the input x is greater than 0, x is output; when the input x is less than or equal to 0, 0 is output; when x = 0, it is non-differentiable, that is, when the input is positive, the original value is maintained, and when the input is negative, 0 is output; however, the ReLU activation function is used after the convolutional layer to increase the non-linearity of the network and promote the model to learn complex patterns.

[0145] S33. The pooling conversion unit is based on the max pooling calculation formula

[0146] "Z(I,j) = max(X[i*Ps(i + 1)*Ps,j*Ps(j + 1)*Ps])

[0147] , Z(I, j) is a pixel value in the output feature map after pooling. I and j are the positions in the output feature map respectively. max is to obtain the maximum value among all pixel values in the input window. X is the input feature map, and Ps is the window size of the pooling operation. "Obtain the maximum pixel value after pooling and enter the fully connected layer to retain important feature information and pass it to the fully connected layer unit;

[0148] Further explanation, calculating the average value of all values in each area through the above maximum pooling calculation formula as the output can retain more information in the feature map, which helps to improve the accuracy of the model, retains more detailed information, and is used in the deeper layers of the network to ensure the integrity of information; before the maximum pooling, obtain the maximum pixel value after average pooling according to the average pooling calculation formula "Z(I, j) = mean(X[i*Ps(i + 1)*Ps, j*Ps(j + 1)*Ps])". Z(i, j) is a certain pixel value in the output feature map after pooling, and mean is to obtain the average value among all pixel values in the input window. According to this formula, select the maximum value from each area of the input feature map as the output, retain the most significant features in the feature map, reduce the size of the feature map, and lose some useful information. Since the maximum value of each area is concerned, the grain size feature is retained, and the features with better classification recognition are selected.

[0149] S34. The fully connected layer unit obtains the output data of the connection layer according to the fully connected layer calculation formula "y = f(∑(Wn*Xn) + b)", where y is the output result, f is the activation function, Wn is the weight of the nth input feature, Xn is the n input features, n is the dimension of the input features, and b is the bias, and enters the output layer for subsequent output calculation and passes it to the normalized output unit;

[0150] Further explanation, the data after pooling is processed by the fully connected layer according to the fully connected layer calculation formula "y = f(∑(Wn*Xn) + b)" to obtain the output value after convolution. In the formula, y is the output result, f is the activation function, Wn is the weight of the nth input feature, Xn is the n input features, n is the dimension of the input features, and b is the bias; the "y = f(∑(Wn*Xn) + b)" is deduced from "y = f(W*X + b)" to "y = f(W 1 *X 1 +W 2 *X 2 +…+Wn*Xn + b)", where W is the weight and Xn is the input feature; through the fully connected layer, the features extracted by the convolutional layer and the pooling layer are integrated and classified, so that each neuron is connected to all neurons in the previous layer, and the output is generated through weighted summation and non-linear activation function.

[0151] S35. The normalization output unit obtains the output of Batch Norm according to the Batch Norm layer calculation formula "h = φ(Bn(Wx + b)), where h is the output of the fully connected layer, φ is the activation function, Bn is the operator of batch normalization, W is the weight parameter, X is the input of the fully connected layer, and b is the bias parameter", and passes it to the output conversion unit;

[0152] For further explanation, the formula "h = φ(Bn(Wx + b))" is derived from "Bn(x) = γ⊙[(X - μb) / σb] + β". In the formula, μb is the mean, σb is the standard deviation, γ is the stretching parameter, β is the offset parameter, and (X - μb) / σb is the standardized normal distribution. The Batch Norm layer (abbreviated as "BN layer") is placed between the affine transformation and the activation function in the fully connected layer, standardizes the output values of each layer in the network, makes them more conform to the normal distribution, can accelerate the training speed of the neural network, and helps prevent gradient disappearance and gradient explosion. The BN layer reduces the correlation between input data by standardizing each mini-batch of data, making the mean of each sample 0 and the variance 1, thus reducing the problem of internal covariate shift, helping to accelerate the training process, improving the stability and generalization ability of the model, and avoiding the problem of reducing the representation ability of the neural network.

[0153] S36. The output conversion unit obtains the size of the output value after convolution according to the output layer conversion calculation formula "N = (P - F + 2C) / S + 1, where N is the size of the output after convolution, P is the size of the input before convolution, F is the size of the convolution kernel, C is the number of layers of adding 0 around the image, and S is the step size", for subsequent model analysis and training, and passes it to the backpropagation unit;

[0154] For further explanation, the output size N is the size of the feature map after the convolution operation, and the input size P is the width or height of the input image or feature map. The padding size P in the height direction is obtained by calculating through the above step S21 H , the padding size P in the width direction W , the convolution kernel size F is obtained by calculating through the above step S22. The number of layers C of adding 0 around the image is the number of layers of adding 0 around the input image or feature map and is used to control the size of the output feature map, and is also obtained by the above network construction. The step size S is the distance that the convolution kernel slides on the input image or feature map and is obtained by the above calculation.

[0155] S37. The backpropagation unit obtains the loss function value according to the loss function calculation formula "L = max(0, m + y*(f(x) - b)), where L is the loss function value, m is the hyperparameter, y is the sample label (0 or 1), f(x) is the predicted value of the model, and b is the classification boundary", and performs backpropagation to optimize the performance of the network model.

[0156] Further explanation: In the loss function calculation formula "L = max(0, m + y * (f(x) - b))", when f(x) ≤ b, then L = m + y(f(x) - b), and when f(x) > b, then L = 0; in machine learning or optimization problems, f(x) is an optimization target, b is a threshold, and m and y are parameters; when the target value f(x) is less than or equal to the threshold b, the value of L is equal to m + y(f(x) - b); when the target value f(x) is greater than the threshold b, the value of L is 0; the model is trained using the training set data, and the model parameters are adjusted to minimize the loss function, so that the predicted value of the positive sample is greater than a certain threshold, and the predicted value of the negative sample is less than a certain threshold; then, the convolution output values dw[l] = dz[l] * a[l - 1], db[l] = dz[l], and da[l - 1] = W[l]T * dz[l] are obtained through the backpropagation calculation formula "dz[l] = da[l] * g[l]′(z[l])", where da[l] is the input data, da[1] is the output data, and g[l]′ is the derivative of the sigmoid activation function; the backpropagation of the convolutional layer passes the error term da[l]da[l] to the previous layer through a convolution operation, flips the convolutional kernel and applies it to the error map, calculates the local gradient through the derivative of the activation function, and is implemented through matrix operations. The input and the convolutional kernel are deformed into matrices, and matrix multiplication is performed, that is, the numbers in the convolution window are pulled into a row to form a column vector, and matrix multiplication is performed; the backpropagation of the fully connected layer calculates the gradients of the weights and the biases through the chain rule. For each node, the error term is passed to the nodes of the previous layer through the weight matrix, and the local gradient is calculated through the derivative of the activation function, and is implemented through matrix operations. The error term is multiplied by the output of the current layer and the input of the previous layer to update it in the direction of minimizing the loss function.

[0157] S40. The model evaluation module validates and evaluates the model according to the trained model, adjusts and optimizes the model according to the validation results to improve the performance and accuracy of the model, and transmits it to the parameter determination module;

[0158] Please refer to Figure 12 As shown, the step S40 includes the following steps:

[0159] S41. The classification and marking unit classifies and marks the injection mold quality anomaly categories into corresponding data sample categories for subsequent machine learning of the model, and transmits it to the learning and training unit;

[0160] Further explanation: Classify according to various quality anomalies in injection molding molds: "Mold cavity wear" is category 1, "Uneven mold temperature" is category 2, "Mold leakage" is category 3, "Difficulty in mold opening and closing" is category 4, "Inappropriate plastic material" is category 5, "Poor appearance of injection molded parts" is category 6, "Color difference on the surface of injection molded parts" is category 7, "Dimension does not match" is category 8; When category 1 is the positive sample, the remaining categories are negative samples, that is, the marked data is (1,0,0,0,0,0,0,0), when category 2 is the positive sample, the remaining categories are negative samples, that is, the marked data is (0,1,0,0,0,0,0,0), when category 3 is the positive sample, the remaining categories are negative samples, that is, the marked data is (0,0,1,0,0,0,0,0), when category 4 is the positive sample, the remaining categories are negative samples, that is, the marked data is (0,0,0,1,0,0,0,0), when category 5 is the positive sample, the remaining categories are negative samples, that is, the marked data is (0,0,0,0,1,0,0,0), when category 6 is the positive sample, the remaining categories are negative samples, that is, the marked data is (0,0,0,0,0,1,0,0), when category 7 is the positive sample, the remaining categories are negative samples, that is, the marked data is (0,0,0,0,0,1,0,0), when category 8 is the positive sample, the remaining categories are negative samples, that is, the marked data is (0,0,0,0,0,1,0,0); The above eight anomaly categories are relatively typical anomaly classifications for injection mold design, including but not limited to the above eight categories. One or more of these categories can also be reclassified into more categories, and more detailed classification can be carried out according to the actual application results. The above classification is not fixed. After each category change, the model is retrained to meet the actual needs of the model.

[0161] S42. The learning and training unit determines the training model according to the gradient descent data, and inputs various mold quality data in the training set into the model for simulation training to ensure the accuracy of model recognition, and transmits it to the model evaluation unit;

[0162] Furthermore, by continuously adjusting the model parameters to minimize the loss function, the prediction accuracy of the model is improved to ensure that the model can accurately identify the adaptability of the injection mold design for applications in different environments; the model is trained with the data set in the training set to fit the data analysis law, that is, various learning parameters such as the weights and biases of the model are determined; and the data set in the validation set is used to adjust the model parameters and hyperparameters during the model training process to optimize the model performance, avoid overfitting, select the model, but does not participate in the determination of the learning parameters, and selects the model parameters and hyperparameters with smaller model errors; then, after the model training is completed, the data set in the test set is used to evaluate the generalization ability of the model on unknown data, which is used once after training to evaluate the effect of the final model, does not participate in the learning parameter process, nor in the hyperparameter selection process; the injection mold quality data includes mold processing data, plastic material data, molding process parameters, inspection data, etc.; the mold processing data includes injection molding machine parameters, mold design parameters, mold material data, mold process data, mold equipment operation parameters, mold inspection data; the mold processing data includes various mold material lists, mold part drawings and assembly drawings, injection molded part drawings, processing technologies for each process, mold manufacturing equipment operation parameters, mold inspection data, etc.; the plastic material data includes material lists, incoming material inspection data, mechanical property parameters of the material, etc.; the molding process parameters include barrel temperature, nozzle temperature, mold temperature, injection pressure, holding pressure, back pressure, injection time, holding time, etc.; the inspection data includes appearance quality, color difference, dimensional tolerance.

[0163] S43. The model evaluation unit obtains the comprehensive evaluation score F according to the model evaluation score calculation formula "F = 2(Ac * Re) / (Ac + Re), where F is the evaluation score of the trained model, Ac is the precision rate, and Re is the recall rate", and transmits it to the processing center.

[0164] Furthermore, Ac and Re are respectively calculated through "Ac = (Tp + Tn) / (Tp + Fp + Fn + Tn), Re = Tp / (Tp + Fn)", where Tp is the number of true positive samples, Fp is the number of actually false positive samples, Fn is the number of false negative samples, and Tn is the number of true negative samples; after the model training is completed, tests are conducted to verify the accuracy and reliability of the model, and problems that may be found in the model during the test process are optimized and adjusted; the evaluation score of the trained model, that is, the F value, is the harmonic mean of the precision rate and the recall rate, which can evaluate the performance of the classification model, and its value range is from 0 to 1, where 1 represents the best performance and 0 represents the worst performance. If the F value is higher, the prediction effect of the model is better, and vice versa. Therefore, if the F value is close to 1, it indicates that the model performs well in maintaining the balance between the precision rate and the recall rate; adjust the hyperparameters according to the performance of the model such as the learning rate and the regularization coefficient to optimize the model performance.

[0165] S44. The processing center compares the actual evaluation score of the model with the standard evaluation score of the model stored in the memory. If it meets the standard, it is the predetermined injection mold process parameters. If it does not meet the standard, it is transmitted to the alarm and notifies to continue training until it meets the standard.

[0166] Further explanation, the best test evaluation score standard of the injection mold quality abnormality category recognition training model is greater than 0.8, which is specifically determined by factors such as the selection of mold steel, the rationality of mold structure, mold process arrangement and machining accuracy, the quality of standard parts, the collision of upper and lower molds (flying mold), the polishing / texture of the cavity, and the characteristics of plastics (type, molecular weight, melt index, etc.), the tonnage and injection speed of the molding machine, the molding conditions (barrel temperature, mold temperature, injection pressure, holding time, etc.), and the process parameters (injection volume, clamping force, molding time, etc.). The injection mold process parameters include mold manufacturing equipment operation parameters, injection molding process parameters, detection processes, etc. Among them, the mold manufacturing equipment operation parameters: the air pressure, power supply status of the mold manufacturing automatic production line, the specifications and quantities of equipment (such as injection molding machines, machining centers, robots, etc.), the process balance rate, water, electricity and gas consumption, servo system parameters, machining center parameters, etc.; the camera resolution, frame rate and exposure of the mold automatic detection equipment, the monitoring area and warning conditions, the automatic adjustment parameter range and warning mechanism, speed gain (servo system parameters), etc. These process parameters are automatically matched and applied in the trial mold verification by the trained model to avoid the occurrence of various quality abnormalities.

[0167] S50. The parameter determination module conducts trial mold verification based on the molding process parameters predicted by the trained model to obtain the best injection mold process parameters, so as to ensure the quality of subsequent molding trial molds and transmit them to the determination scheme module;

[0168] Please refer to Figure 13 As shown, the step S50 includes the following steps:

[0169] S51. The mold size unit calculates the mold size according to the mold size calculation formula "Dm 1 =(D M +Im+Es)*(1+Za), Dm 2 =D M +Im+Es+Bh+Cg, Dm 3 =D M3 *(1+So)-Cv, Dm 1 、Dm 2 、Dm 3 are the outer shape, structure, and radial size of the cavity of the mold respectively, D M is the size of the injection product, D M3The radial limit dimension of the injection molded part is D, the inner margin of the mold is Im, the shrinkage at both ends of the injection molded part is Es, the machining allowance of the mold is Za, the height of the mold boss is Bh, the mold clamping gap is Cg, and the shrinkage rate of the plastic material is So. The mold correction value Cv is used to obtain the dimensional data of the injection mold for subsequent injection mold design marking and transfer it to the forming pressure;

[0170] For further explanation, the dimension D of the injection molded product M and the radial limit dimension D of the plastic part M3 are obtained from the dimensions of the injection molded part or sample provided by the customer, where D M3 is the maximum or minimum allowable value of the radial dimension of the injection molded part; the inner margin of the mold Im and the height of the mold boss Bh are determined according to the design requirements of the customer's injection molded product and the mold manufacturing standards. Here, Im is the distance between the net dimension of the injection molded product and the inner cavity dimension of the mold, and Bh is the protruding part inside the mold; the shrinkage at both ends of the injection molded part Es is determined by the shrinkage rate of the plastic material and the shape of the injection molded part. Specifically, it is calculated according to the formula "Es = D M + D M * So + D M * So2"; the mold clamping gap Cg is the gap reserved for tight fitting when the mold is clamped, and is determined by the manufacturing accuracy of the mold, the fluidity of the plastic material, and the clamping force of the injection molding machine; the shrinkage rate So of the plastic material is obtained from the material supplier or the material database; the machining allowance Za of the mold is the gap between the mold and the injection molded part, and is a reserved dimension set to ensure the machining accuracy and service life of the mold (due to forming wear). It is calculated according to the formula "2Za = (di - 1) - di". In the formula, 2Za is the bilateral machining allowance remaining after removing the metal layer with a thickness of Za from both sides of the part, Za is the allowance for this process, di - 1 is the basic dimension of the previous process, and di is the basic dimension of this process, which helps to ensure the machining accuracy and surface quality of the part, and at the same time is beneficial to improving the machining efficiency and reducing the machining cost, and is affected by factors such as the dimensional tolerance, shape and position tolerance, surface roughness and defect layer of the previous process, and the installation error of this process; the mold correction value Cv is the value for adjusting the cavity dimension of the mold to compensate for the deviations in the design, manufacturing or use process, and is determined by the specific processing conditions and mold structure, and is also affected by factors such as the draft angle, mold manufacturing accuracy, selection of mold material, determination of the parting line and the mold opening direction, design of the cooling system, and layout of the runner system. Therefore, these factors must be considered to improve the mold design scheme.

[0171] S52. The forming pressure unit calculates according to the injection molding machine forming pressure formula "Pi = P * Sa / [π * (d / 2) 2 , P F= Sm * Pv / 1000, where Pi is the injection pressure of the injection molding machine Pi (kg / cm 2 ), P is the pump pressure of the injection molding machine (kg / cm 2 ), Sa is the effective area of the injection cylinder of the injection molding machine (cm 2 ), d is the diameter of the screw of the injection molding machine (cm), P F is the clamping force of the injection molding machine (T), Sm is the projected area of the mold cavity of the injection molding machine (cm 2 ), Pv is the filling pressure of the injection molding machine (kg / cm 2 )” are used to obtain the molding injection pressure and the clamping pressure respectively for subsequent mold forming debugging and are transmitted to the injection weight unit;

[0172] Further explanation, the formula for calculating the injection pressure of the injection molding is obtained by substituting “So = π * (d / 2) 2 ” into “Pi = P * Sa / So”. In the formula, So is the cross-sectional area of the screw of the injection molding machine (cm 2 ); the pump pressure P of the injection molding machine, the effective area Sa of the injection cylinder of the injection molding machine, and the diameter d of the screw of the injection molding machine are all obtained by querying the design parameters of the injection molding machine manufacturer; when the injection pressure is output, the injection speed is matched and obtained by the formula “υ = Q / A 2 ”. In the formula, υ is the injection speed of the injection molding machine (cm / s), Q is the discharge volume of the pump of the injection molding machine (cm 3 / s), Sa is the effective area of the injection cylinder of the injection molding machine (cm 2 ); the discharge volume Q of the pump of the injection molding machine and the effective area Sa of the injection cylinder of the injection molding machine are all obtained by querying the technical specification of the injection molding machine; the effective area Sa of the injection cylinder of the injection molding machine is the effective area that can directly act on the plastic melt. The plastic flow rate when the molten plastic fluid flows through the mold runner is obtained by the formula “Q = S A * υv”. In the formula, Q is the flow rate of the plastic flowing through the pipeline per unit time (m 3 / s), S A is the cross-sectional area of the plastic fluid flowing through the mold pipeline (m 2 ), υv is the fluid velocity of the fluid in the pipeline (m / s); the clamping force P of the injection molding machine FThe clamping force exerted by the mold clamping mechanism of the injection molding machine on the mold to prevent the mold from separating due to the pressure of the molten plastic during the injection molding process is affected by factors such as the size and shape of the injection molded part and the plastic material used; the projected area Sm of the injection molding machine mold cavity is the projected area of the mold cavity in the direction perpendicular to the mold clamping direction to ensure that the mold will not open due to the pressure of the plastic during the injection molding process and is obtained by measuring the dimensions of the mold cavity; the filling pressure Pv of the injection molding machine is the pressure exerted by the injection molding machine when injecting the molten plastic into the mold cavity, which is determined by the fluidity of the plastic material and is obtained through the plastic material technical specification or injection molding historical experience data, generally between 150 - 350 KG / CM 2 For materials with good fluidity, a lower value is taken, and for materials with poor fluidity, a higher value is taken.

[0173] S53. The injection weight unit obtains the injection weight of the injection molding machine during molding according to the injection molding machine molding injection weight calculation formula "Wv = (π * (d / 2) 2 * Lt) * η * δ, where Wv is the injection weight (g) of the injection molding machine, d is the diameter (cm) of the injection molding machine screw, Lt is the injection stroke (cm) of the injection molding machine, η is the specific gravity of the plastic material (g / cm 3 ), and δ is the efficiency of the injection molding machine" for subsequent injection mold molding debugging and transmits it to the ejector pin strength unit;

[0174] Further explanation, the injection molding injection weight calculation formula "is obtained by substituting 'V = π * (d / 2) 2 * Lt' into 'Wv = V * η * δ'. In the formula, V is the injection volume (cm 3 ); the screw diameter d is obtained by querying the technical specification of the injection molding machine; the injection stroke Lt is the distance that the screw moves during the injection process and together with the screw diameter determines the size of the injection volume and is obtained by detecting through the set sensor; the specific gravity η of the plastic material is the ratio of the mass of the substance to its volume under specific conditions such as temperature and pressure, that is, the mass of plastic per unit volume, and is obtained by querying the specification of the plastic material provided by the supplier; the efficiency δ of the injection molding machine is the energy conversion efficiency during the injection molding process, that is, the ratio of the actual injected plastic weight to the theoretical injected weight, which is determined by factors such as the injection molding machine, screw, mold, and various energy losses during the injection process, and is determined according to historical experience data or the performance data of the injection molding machine, and varies due to different injection conditions and materials, and generally takes an ideal value of about 0.95.

[0175] S54. The ejector pin strength unit calculates according to the injection mold ejector pin strength calculation formula "F = π 2 * λ * (π * (d o / 2) 2 ) * E * r 2 / L 2, F is the buckling load (kgf) of the mold ejector pin, λ is the support condition constant of the mold ejector pin, d o is the diameter (cm) of the mold ejector pin, E is the longitudinal elastic modulus (kgf / cm 2 ) of the mold ejector pin, r is the sectional radius of inertia (cm) of the mold ejector pin, and L is the length (cm) of the mold ejector pin” to obtain the strength of the mold ejector pin for subsequent injection mold forming debugging and transfer it to the mold manufacturing unit;

[0176] Furthermore, the buckling load F of the mold ejector pin is the maximum load that the ejector pin can withstand when subjected to pressure; the support condition constant λ of the mold ejector pin is the influence of the support method of the ejector pin on its buckling load. Different support conditions such as fixed at both ends and hinged at both ends correspond to different λ values, and the values are taken differently according to different support situations. If it is a straight rod, then λ is 4, and if it is a stepped rod, then λ is 2.05, which can be obtained specifically by referring to relevant mechanics manuals or textbooks; the diameter d of the mold ejector pin o is obtained from the mold design drawing, and the cross-sectional area of the ejector pin is calculated according to (π*(do / 2) 2 ), and its size directly affects the strength and load-bearing capacity of the ejector pin; the longitudinal elastic modulus E of the mold ejector pin is the ability of the mold ejector pin material to resist deformation when subjected to longitudinal force. The larger the elastic modulus E, the more difficult it is for the material to deform, and the stronger the load-bearing capacity of the ejector pin. It is obtained through material mechanics experiments or by referring to material manuals, and generally takes a value of 21000 kgf / mm 2 ; the sectional radius of inertia r of the ejector pin is the bending resistance ability of the ejector pin cross-sectional shape. If the radius of inertia is larger, the cross-section is less likely to bend. If it is a circular cross-section, then r is equal to one-fourth of the ejector pin diameter. If it is other shapes, it is calculated by the formula r = sqrt[I / (π*(d o / 2) 2 )]; the length L of the mold ejector pin is obtained from the mold design parameters. If the length of the ejector pin is longer, the ejector pin is more likely to bend when subjected to the same force, so its buckling load will also decrease accordingly. By reasonably selecting factors such as support conditions, cross-sectional shape and size, and material, it is ensured that the ejector pin has sufficient strength and stability during the injection process, thus avoiding failure or damage caused by buckling.

[0177] S55. The mold manufacturing unit designs drawings according to the predetermined mold manufacturing process parameters, manufactures the mold through the mold manufacturing equipment, and obtains the mold processing quality information to ensure the quality of subsequent forming trial production and transfer it to the forming debugging unit;

[0178] Further explanation: Calculate and input the obtained data into the model according to the customer order information and the above steps S51 - S54 to match the corresponding injection mold process parameters, avoiding various mold quality problems caused by human calculation and analysis errors and incomplete data judgment. Conduct a detailed analysis of the dimensions, shapes, materials, wall thicknesses, etc. of the injection molded parts to be injected and check for system data matching to ensure the rationality and feasibility of the mold design. Then, design a reasonable mold structure according to the product characteristics and use drawing software such as CAD to draw detailed mold drawings, including 3D models and 2D drawings, for subsequent mold processing; Select the required mold materials such as mold steel, guide pillars, and guide sleeves according to the design drawings to determine the accuracy, durability, and production cost of the mold; Conduct preliminary rough machining on the mold materials through mold manufacturing equipment to form the basic mold shape, and then perform fine machining to ensure the accuracy and surface quality of the mold; Then, inspect the mold accessories. If the dimensional accuracy and surface quality of the mold parts are all qualified, assemble the mold and check the mating conditions of each part. If they do not match, perform appropriate trimming and flashing to make the mold cavity of the mold core fully cooperate.

[0179] S56. The forming and debugging unit conducts forming and debugging through an injection molding machine according to the predetermined forming process parameters and obtains the quality information of the injection molded parts to ensure the accuracy of the mold design and transmits it to the processing center;

[0180] Further explanation: Install the assembled mold on the predetermined injection molding machine according to the predetermined machine tonnage, injection speed, barrel temperature, mold temperature, injection pressure, holding time, and other forming conditions, injection volume, clamping force, forming time, and other forming process parameters; Add the dry granular plastic raw materials into the barrel of the injection molding machine to avoid generating bubbles or defects during the injection molding process; Gradually heat and melt the raw materials through the barrel heating device, and stir, compact, and convey the plastic through the rotation of the screw in the injection molding machine to uniformly plasticize it into a molten state to ensure the plasticization quality of the plastic; When the plastic is completely plasticized, push the molten plastic forward through the screw and inject it into the closed mold cavity at high pressure and high speed through the nozzle of the injection molding machine, using parameters such as injection pressure, speed, and holding time to ensure the density, dimensional accuracy, and appearance quality of the injection molded product; The molten plastic injected into the mold gradually cools and solidifies under the action of the mold cooling system to form the shape of the product to ensure that the injection molded product is fully shaped and avoid deformation. After cooling is completed, the mold automatically opens and the formed injection molded product is ejected from the mold through the ejection mechanism; Then, obtain the appearance quality and dimensional deviation of the injection molded parts through a high-definition camera and a coordinate measuring instrument respectively, and compare with the calculation results of the above steps S51 - S54 according to the test results and match with the model to make necessary adjustments and optimizations to the mold to ensure that the injection molded product meets the customer quality requirements.

[0181] S57. The processing center compares the quality information of the injection mold with the quality standard of the injection mold stored in the memory: if it meets the standard, it is set as the formal injection mold process parameters; if it does not meet the standard, it is passed to the alarm and the parameter adjustment and rework are notified.

[0182] Further explanation, the quality standard of the injection mold includes: 1) Appearance requirements: The gloss of the surface of the injection molded part is normal, without color difference, flash, crack, unevenness, and without defects such as bubbles, flow marks, sink marks, scars, burn marks, color mixing, water entry marks, fog spots, concave points, convex points, flower marks, shrinkage, spots, weld lines, etc., to improve the aesthetics and market competitiveness; 2) Dimension requirements: The dimension requirements of injection molded parts with high precision requirements in industries such as automotive and electronics include that the dimension deviation, roundness, straightness, perpendicularity, parallelism, angle, etc. meet the standards. The dimensional tolerance of precision injection molded parts is extremely small, up to ±0.01mm or even smaller, to ensure the assembly accuracy and service performance of the product; 3) Material requirements: The strength, hardness, and wear resistance of the plastic material for mechanical part profiles must meet the regulations, and it is non-toxic and harmless, meeting the environmental protection requirements, to ensure the safety and reliability of the injection molded product; 4) Physical property requirements: The physical properties such as strength, hardness, toughness, heat resistance, cold resistance, and corrosion resistance of the injection molded product meet the industry and customer requirements according to the use environment and functional requirements of the injection molded product, to ensure that the injection molded product has sufficient stability and durability during use.

[0183] S58. The anomaly recognition unit matches the real-time obtained quality inspection image of the injection mold with the corresponding image in the trained convolutional neural network model and confirms its anomaly category for subsequent molding improvement and parameter adjustment.

[0184] Further explanation: Input the image into the model, use the trained model to predict the new monitoring image, and judge the type of abnormal quality of the injection mold according to the output result. If the judgment of the abnormal quality of the injection mold by the model reaches a certain threshold, trigger the warning system and take corresponding solutions in a timely manner; the types of abnormal quality of the injection mold include: 1) Mold cavity wear: Since the mold is prone to wear due to reasons such as material friction and high-temperature thermal expansion and contraction during the molding process, resulting in problems such as inaccurate dimensions and uneven surfaces of the injection-molded products. For subsequent model training and learning, mold repair, mold replacement, the use of highly wear-resistant materials, and regular inspection and maintenance of the mold should be carried out; 2) Uneven mold temperature: Due to improper control of the injection molding machine, poor mold surface treatment, etc., the mold temperature is uneven, resulting in problems such as deformation and cracking of the injection-molded products. For subsequent model training and learning, the parameters of the injection molding machine should be adjusted, the mold surface treatment should be improved, and a temperature control system should be installed to ensure uniform mold temperature for subsequent model training and learning; 3) Mold water leakage: Due to water leakage in the mold cooling water channel, defects such as water stains and bubbles appear on the injection-molded parts. For subsequent model training and learning, seals should be inspected and replaced, broken water channels should be repaired, and the processing accuracy and sealing performance of the cooling water channel should be improved; 4) Difficult mold opening and closing: Due to wear of the mold guiding mechanism, the mold gets stuck and has excessive resistance during the mold opening and closing process. For subsequent model training and learning, the guide pillars and guide sleeves should be replaced to ensure their tight fit for subsequent model training and learning, and the processing accuracy and sealing performance should be improved during the mold design and manufacturing process; 5) Unsuitable plastic material: Due to the selection of inappropriate plastic materials, problems such as unstable quality and easy deformation of the injection-molded products occur. For subsequent model training and learning, the plastic material should be replaced, the parameters of the injection molding machine should be adjusted, and the mold design should be improved to ensure the selection of suitable plastic materials for subsequent model training and learning; 6) Poor appearance of the injection-molded part: Due to abnormal mold design (such as too small mold gate runner, poor mold exhaust, etc.), unreasonable molding parameters, or abnormal machine conditions, or poor material quality, appearance abnormalities such as shrinkage of the injection-molded product occur. For subsequent model training and learning, the water inlet position should be adjusted, the cooling water channel design should be optimized, the mold temperature should be ensured to be uniform, etc., or the molding parameters should be adjusted, or the machine should be replaced, or high-quality materials should be replaced; 7) Color difference on the surface of the injection-molded part: Due to unstable mold temperature, uneven mixing of raw materials, different raw material batches, or insufficient drying of raw materials, the color of the injection-molded part is different from the standard color sample, or the color is uneven (color mixing), or the gloss is poor (dark color). For subsequent model training and learning, high-quality plastic materials should be replaced and the mold temperature should be adjusted; 8) Dimension mismatch: Due to unreasonable mold design or improper control of injection molding conditions, the dimensions of the injection-molded part are inconsistent with the predetermined dimensions. The mold design or molding parameters should be optimized.

[0185] S60. The solution determination module confirms the shrinkage rate according to the quality information of the injection-molded part qualified by the trial mold verification, improves the mold design drawing, and inspects the appearance quality of the injection mold to ensure that the injection mold design meets the customer quality requirements.

[0186] Please refer to Figure 14 as shown, the step S60 includes the following steps:

[0187] S61. The shrinkage confirmation unit obtains the shrinkage rate of the injection molded part according to the injection molded part shrinkage rate calculation formula "Sr = [-D M +sqrt(4D M *Dm - 3D M 2 )] / 2D M , where Sr is the shrinkage rate of the injection molded part, Dm is the mold size, and D M is the injection molded part size" to confirm the accuracy of the injection molded part and transmit it to the appearance detection unit;

[0188] Further explanation, the injection molded part shrinkage rate calculation formula is derived through "Dm = D M +D M *Sr + D M *Sr 2 ", that is, the percentage of dimensional shrinkage generated by the plastic during the molding process due to cooling and solidification. When the shrinkage rate is small or the difference between the mold size and the injection molded part size is small, the formula "Sr = (Dm - D M ) / Dm * 100%" is used to replace the above formula to calculate the shrinkage rate. Since the shrinkage rate Sr is usually a positive number and less than 1 (less than 100% when expressed as a percentage) in actual applications, if the calculated Sr value is not between 0 and 1 (or between 0% and 100%, expressed as a percentage), the input data is incorrect or affected by other factors such as the type of plastic, molding conditions (temperature, pressure, etc.), and mold structure, resulting in data errors; the mold size Dm and the injection molded part size D M are both the mold size and the injection molded part size after the trial mold of this injection mold is verified to be qualified, and are automatically obtained from the qualified inspection data. Therefore, substituting the mold size Dm and the injection molded part size D M into the injection molded part shrinkage rate calculation formula respectively, the shrinkage rate of this injection molded part is obtained, and then compared with the shrinkage rate of this injection material. If they are consistent, the process parameters of this injection mold are accurate, and thus this design scheme is also accurate; if the actual shrinkage rate Sr of the injection molded part is close to or equal to the shrinkage rate So of the plastic material, it proves that the molding quality is completely qualified and proceeds to step S62 to determine the design scheme. If the actual shrinkage rate Sr of the injection molded part is inconsistent with the shrinkage rate So of the plastic material, it proves that the molding quality is unqualified and is passed to step S56 to re-adjust the mold until they are consistent.

[0189] S62. The drawing determination unit improves the injection mold design drawings and their corresponding design information according to the parameters after the trial mold debugging to determine the final mold design scheme and transmit it to the appearance detection unit;

[0190] Further explanation: The mold design plan includes: 1) Design task sheet: the geometric shape, usage requirements, raw materials, and molding processability of the customer's injection molded parts to ensure design feasibility, as well as the model and specifications of the matching injection molding machine and the mold bill of materials required for mold manufacturing; 2) Molding process card: general product information such as the sketch, weight, and wall thickness of the injection molded parts, plastic material information such as product name, model, and color, main technical parameters of the injection molding machine such as the relevant dimensions between the injection molding machine and the installed mold, and the screw type, injection molding machine pressure and stroke, and injection molding conditions such as temperature, pressure, and speed; 3) Mold structure design: the number of cavities can meet the requirements of the maximum injection volume, clamping force, product accuracy, and economy, the parting surface can ensure a simple mold structure, easy parting, and no impact on the appearance and use of the plastic part, the distribution of the cavities adopts a balanced arrangement, the design of the gating system such as the sprue, runner, and gate can ensure that the plastic melt can smoothly enter the cavity, the demolding method can adapt to the position of the injection molded part in the mold, the temperature control system can meet the requirements of the injection process for the mold temperature, the exhaust system discharges the air in the cavity and the gas volatilized from the plastic itself during molding out of the mold, and the main dimensions of the injection mold such as the working dimensions of the molding parts and the side wall thickness of the mold cavity; 4) Standard mold base: the main dimensions of the standard mold base are calculated based on the size of the injection molded part; 5) Mold drawings: the general assembly drawing of the mold, including the structure of the mold forming part, gating system, and exhaust system, and the part drawings marked with serial numbers and detailed lists; 6) Proofreader / Reviewer: The proofreader and reviewer of the mold drawings can ensure the accuracy and feasibility of the design.

[0191] S63. The appearance detection unit obtains images or videos of the injection mold through a high-definition camera and converts them into recognizable appearance quality information for subsequent packaging and shipping and mass production of molding, and transmits them to the processing center;

[0192] Furthermore, through the high-definition camera set, 360-degree shooting is carried out to capture images of different position points of the injection mold / parts from multiple different angles, so as to identify different position points of the injection mold / parts from different angles, and then be able to more accurately and comprehensively identify and obtain the position points and their shapes of the injection mold / parts. Or, based on the images of the injection mold / parts captured from different angles, the injection mold / parts are identified by multi-angle matching, and then the appearance quality information of the injection mold / parts is determined, including: whether the mold appearance is smooth and flat, whether there are obvious scratches, sand holes, cracks and other defects, whether the surface plating or coating is uniform and firm, without peeling or flaking phenomena, whether the marks such as mold numbers, specifications, production dates are clear and accurate; whether the content of the mold nameplate is complete, whether the characters are clear, whether the arrangement is neat, whether it is fixed on the mold feet near the template and the reference angle, whether the fixation is reliable and not easy to peel off, whether there are inlet and outlet marks on the cooling water nozzles, and whether the English characters and numbers of the marks are clear, beautiful and neat; whether the surface of the mold base has defects such as pits, rust, redundant lifting rings, inlet and outlet water vapor, oil holes, etc.; according to specific injection molding requirements and algorithm requirements, it is also necessary to collect other types of image data such as images under different lighting conditions and images at different angles to improve the accuracy and robustness of the algorithm.

[0193] S64. The processing center compares the quality information of the injection mold with the quality standard of the injection mold stored in the memory: if it meets the standard, it notifies the customer that mass production can be carried out; if it does not meet the standard, it is passed to the alarm and rework is notified.

[0194] For further clarification, in order for subsequent model training and learning of the injection mold quality standards, in addition to the injection mold quality standards described in step S57 above, it also includes: 1) Mold appearance: The appearance of the injection mold is smooth and flat, without obvious scratches, sand holes, cracks and other defects. The surface plating or coating should be uniform and firm, without peeling or flaking. The mold identification should be clear and accurate, including information such as mold number, specification, production date, etc.; 2) Mold nameplate: The content of the mold nameplate is complete, the characters are clear, neatly arranged, fixed on the mold feet near the template and the reference corner, fixed reliably and not easily peeled off. The cooling water nozzles have inlet and outlet marks, and the marked English characters and numbers should be clear, beautiful and neat; 3) Die set surface: The die set surface has no pits, rust, unnecessary lifting rings, incoming and outgoing water vapor, oil holes and other defects; 4) Mold structure: The structure of the mold is reasonable and firm. The connection between each component should be tight and reliable. The mold opening and closing actions are smooth, without jamming or abnormal noise. The guiding mechanism and positioning mechanism of the mold should be accurate to ensure the accuracy of the mold when closing; 5) Cooling system: The cooling system can effectively control the mold temperature, make the temperature of the mold cavity surface uniform. There are certain requirements for the diameter, spacing of the cooling pipes and the distance from the cavity surface to ensure the cooling efficiency; 6) Ejection system: The ejection force of the ejection system is evenly distributed to avoid deformation or damage of the injection molded product. The number, position and diameter of the ejector pins should be determined according to the shape, size and material of the injection molded product.

[0195] For further clarification, the above steps are shown in sequence according to the arrows. However, these steps are not necessarily executed in the order indicated by the arrows, unless there is a clear description in this article. The execution of these steps has no strict order limit. These steps can also be executed according to other orders. Moreover, some steps can include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0196] Further explanation: The present invention is described in accordance with the content implemented by the software program of a control system for AI-based injection mold design. For each AI-based rapid injection mold design method, it is divided into several modules or units to implement the software program instructions generated by each step. The software program instructions include the above-mentioned AI-based rapid injection mold design method.

[0197] The system of the present invention further includes a computer-readable storage medium containing a memory; the memory stores a computer program, and when each functional module executes the computer program, it realizes the steps of the AI-based rapid injection mold design method described in any one of the above; a computer program is stored on the computer-readable storage medium, and when the computer program is executed by each functional module, it realizes the steps of the AI-based rapid injection mold design method described in any one of the above.

[0198] Further explanation: The computer-readable storage medium includes a memory, which can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the programs and instructions corresponding to the AI-based rapid injection mold design method in the present invention, including the information transfer instructions of each module; the memory executes various functional applications and data processing of each module by running the stored non-volatile software programs and instructions, that is, it realizes the AI-based rapid injection mold design method of the above process embodiment; one or more units are stored in the memory, and when executed by the one or more modules, they execute the AI-based rapid injection mold design method in any of the above process embodiments; the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by one or more modules, it can also be the AI-based rapid injection mold design method in any of the above process embodiments.

[0199] Further explanation, the computer-readable medium can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two above. It can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or component, including but not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. Among them, the computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, carrying computer-readable program code. The propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. It can also be any computer-readable medium other than the computer-readable storage medium. This computer-readable signal medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, device, or component. The program code it contains can be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0200] Further explanation, the computer program is divided into multiple modules / units, and the multiple modules / units are stored in the memory and executed by each of the modules / units to complete the present invention. The multiple modules / units can be a series of computer program instructions that can complete specific functions, and these instructions are used to describe the execution process of the computer program in each of the modules / units.

[0201] Further explanation, the content described in the present invention is written by referring to some parts implemented in the software copyrights such as "Plastic Mold Process Analysis Software (Registration Number: 2018SR259849)", "Injection Molding Module Design System (Registration Number: 2018SR260868)", "Intelligent System for Quality Inspection and Analysis of Plastic Molds (Registration Number: 2020SR1764697)", "Plastic Mold Molding Control Terminal System (Registration Number: 2020SR1764722)", etc., which were successively applied by our company starting from April 2018.

[0202] The present invention also provides an AI-based injection mold design control device, which is implemented using the above-mentioned AI-based rapid injection mold design method.

[0203] Further explanation: The control device is composed of the above-mentioned associated detection devices or apparatuses. According to the needs of injection mold enterprises, it can be made into a complete set of mold automatic production lines together with injection mold manufacturing equipment, or made into unit devices of each functional module belonging to each step or the entire control device, including automatic production lines for mold processing, automatic injection molding machines, automatic appearance size detection equipment, etc. When installed, wireless connection is used to connect each control device to the injection mold manufacturing equipment; the structures of these devices will not be described in detail here.

[0204] For those skilled in the art, it is obvious that this application is not limited to the details of the above-mentioned exemplary embodiments, and without departing from the spirit or basic characteristics of this application, this application can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of this application is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in this application.

[0205] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and not to limit them. Although this application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of this application can be modified or equivalently replaced, and all of them should be included in the protection scope of this application.

Claims

1. A rapid design method for injection molds based on AI, characterized by: The method is applied to an AI-based injection mold design control system, the system comprising an information acquisition module, a model construction module, a model training module, a model evaluation module, a parameter determination module, a solution determination module, a wireless communication module, a memory, an alarm, a processing center, and an intelligent mobile terminal; the information acquisition module, the model construction module, the model training module, the model evaluation module, the parameter determination module, the solution determination module, the wireless communication module, the memory, and the alarm are respectively connected to the processing center; the intelligent mobile terminal is respectively connected to the wireless communication module wireless network within the range of a wireless network or the Internet; the method comprises the following steps: S10. Before debugging, the information acquisition module acquires normal and abnormal historical image data similar to the molding quality of the injection mold and performs preprocessing for subsequent model construction, and passes it to the model construction module; S20, the model building module determines the model architecture, designs the network hierarchy, adds auxiliary layers, and defines model parameters according to the extracted mold feature information to build the model for subsequent model training and verification, and passes it to the model training module; S30, the model training module trains the model according to the collected data and adjusts the model parameters to optimize the model performance to predetermine the injection mold process parameters, ensure the quality stability of the injection mold design, and pass it to the model evaluation module; S40, the model evaluation module performs model verification and evaluation based on the trained model, and adjusts and optimizes the model based on the verification results to improve the performance and accuracy of the model, and passes it to the parameter determination module, including the following steps: S41, the classification and labeling unit classifies the data samples into corresponding data sample categories according to the quality abnormality categories of the injection mold and labels them for the purpose of machine learning of the subsequent model, and transmits them to the learning and training unit; S42, the learning training unit determines the training model according to the gradient descent data, and inputs various mold quality data in the training set into the model for simulation training to ensure the accuracy of model recognition, and transmits it to the model evaluation unit; S43, the model evaluation unit obtains the comprehensive evaluation score F according to the model evaluation score calculation formula "F = 2 (Ac * Re) / (Ac + Re), F is the training model evaluation score, Ac is the precision, Re is the recall rate", and transmits it to the processing center; S44, the processing center compares the actual evaluation score of the model with the evaluation score standard of the model stored in the memory: if it meets the standard, it is the predetermined injection mold process parameter; if it does not meet the standard, it is transmitted to the alarm and notified to continue training; S50, the parameter determination module performs a trial mold verification according to the molding process parameters predicted by the trained model to obtain the best injection mold process parameters to ensure the quality of subsequent molding trial molds, and passes them to the solution determination module; S60, the solution determination module confirms the shrinkage rate based on the quality information of the injection molded parts that have passed the mold trial verification, improves the mold design drawings, and detects the appearance quality of the injection mold to ensure that the injection mold design meets the customer's quality requirements.

2. The AI-based rapid injection mold design method according to claim 1, characterized in that: The system further comprises: the wireless communication module is provided with a wireless network unit, which is responsible for sending and receiving wireless signals and automatically networking with the intelligent mobile terminal within an effective network range; The alarm device compares the actual evaluation score of the model with the evaluation score standard of the model stored in the memory, and automatically sounds an alarm and notifies to continue training if the standard is not met; and compares the mold molding quality information with the injection mold quality standard stored in the memory, and automatically sounds an alarm and notifies to adjust parameters for rework if the standard is not met; and compares the quality information of the mold with the injection mold quality standard stored in the memory, and automatically sounds an alarm and notifies to rework if the standard is not met; The memory is responsible for storing information of the information acquisition module, the model construction module, the model training module, the model evaluation module, the parameter determination module, the solution determination module, the wireless communication module, and the alarm, as well as the model evaluation sub-standard and the injection mold quality standard; The processing center is responsible for the information acquisition module, model construction module, model training module, model evaluation module, parameter determination module, solution determination module, wireless communication module, alarm, and information transmission of the memory. It is the hub center of the system and compares the actual evaluation score of the model with the evaluation score standard of the model stored in the memory: if it meets the standard, it is the predetermined injection mold process parameter; if it does not meet the standard, it is passed to the alarm and notified to continue training; and compares the mold molding quality information with the injection mold quality standard stored in the memory: if it meets the standard, it is set as the formal injection mold process parameter; if it does not meet the standard, it is passed to the alarm and notified to adjust the parameters for rework; and compares the quality information of the mold with the injection mold quality standard stored in the memory: if it meets the standard, the customer is notified that mass production can be carried out; if it does not meet the standard, it is passed to the alarm and notified to rework; The information acquisition module includes a data acquisition unit, a cleaning data unit, an enhancement data unit, an integration data unit, a conversion data unit, a filtering and denoising unit, a grayscale conversion unit, and a feature extraction unit, which is responsible for acquiring and preprocessing normal and abnormal historical image data similar to the molding quality of injection molds, and passing them to the model building module; The model building module includes a network level unit, a network parameter unit, a hierarchical optimization unit, and a data partitioning unit, which are responsible for determining the model architecture, designing the network level structure, adding auxiliary layers, defining model parameters, and building the model according to the extracted feature information, and passing them to the model training module; The model training module includes an input filling unit, a convolution input unit, a pooling conversion unit, a fully connected layer unit, a normalized output unit, an output conversion unit, and a back-propagation unit, which is responsible for training the model according to the collected data and adjusting the model parameters to optimize the model performance to predetermine the injection mold process parameters and pass them to the model evaluation module; The model evaluation module includes a classification labeling unit, a learning training unit, and a model evaluation unit, which is responsible for model verification and evaluation based on the trained model, and adjusting and optimizing the model based on the verification results, and passing it to the parameter determination module; The parameter determination module includes a mold size unit, a molding pressure unit, an injection weight unit, an ejector pin strength unit, a mold making unit, a molding debugging unit, and an abnormality identification unit, which is responsible for performing mold development tests based on the molding process parameters predicted by the trained model to obtain the best injection mold process parameters and pass them to the solution determination module; The solution determination module includes a shrinkage confirmation unit, a drawing determination unit, and an appearance inspection unit, which are responsible for confirming the shrinkage rate according to the qualified injection molded part quality information, improving the mold design drawings, and inspecting the appearance quality of the injection mold to ensure that the injection mold design meets the customer's quality requirements; The step S20 comprises the following steps: S21, the network level unit uses the selected convolutional neural network as the model architecture and customizes and optimizes the network level structure according to the specific application scenarios and requirements for subsequent defect detection and quality identification, and passes it to the network parameter unit; S22, the network parameter unit adjusts the number of network layers, convolution kernel size, step size, padding method and selects the loss function and optimization algorithm according to the network hierarchy structure to ensure the performance and accuracy of the model, and passes it to the hierarchical optimization unit; S23, the hierarchical optimization unit adjusts the size and normalization of the input data by adding a zero padding layer and a normalization layer before and after the convolution layer to improve the model training speed and stability, and passes it to the data partitioning unit; S24. The data partitioning unit divides the data set with key feature vectors into training set, validation set and test set according to the extraction and establishes a convolutional neural network structure for subsequent model training and learning.

3. The AI-based rapid injection mold design method according to claim 1 is characterized in that: The step S10 comprises the following steps: S11. The data acquisition unit obtains various quality data, mold data and historical data of molding process parameters similar to injection mold molding by networking with the industry database and performs preprocessing for subsequent data cleaning, and transmits it to the cleaning data unit; S12, the cleaning data unit cleans the collected mold history data to remove abnormal values, duplicate values ​​or missing values ​​in the data to improve the quality and accuracy of the data, and transmits it to the enhanced data unit; S13, the enhanced data unit increases the quantity and diversity of the injection mold training data through data enhancement methods such as rotation, scaling, flipping, cropping, and color transformation to improve the generalization ability of the model, and passes it to the integrated data unit; S14, the integrated data unit obtains the standardized injection mold data according to the data standardization processing formula "z = (x-μ) / σ, z is the standardized data, x is the original data, μ is the mean value of the data, σ is the standard deviation of the data" to improve the stability of the model, and transmits it to the conversion data unit; S15, the conversion data unit obtains the normalized injection mold data according to the normalization calculation formula "x'=(x-min(x)) / (max(x)-min(x)), x' is the normalized data, x is the original data, min(x) and max(x) are the minimum and maximum values ​​of the data respectively" to improve the model performance, and passes it to the filtering and denoising unit; S16, filtering and denoising unit is calculated according to the Gaussian filtering method" G(x,y) is the two-dimensional Gaussian function pixel, (x,y) is the pixel coordinate, and σ is the standard deviation” The denoised injection mold design image is obtained for grayscale image conversion and passed to the grayscale conversion unit; S17, the grayscale conversion unit obtains the grayscale injection mold design image according to the grayscale calculation formula "f(I,j)=max(R(I,j),G(I,j),B(I,j)), f(I,j) is the grayscale image, R(I,j), G(I,j), B(I,j) are the original images of three colors respectively", so as to extract features in the subsequent process, and transmits it to the feature extraction unit; S18, the feature extraction unit converts the annotated data into feature vectors useful for model training and extracts key features of the geometric shape, surface quality, dimensional accuracy, and demolding slope of the mold and injection molded parts to improve the training speed and performance of the model; The step S50 comprises the following steps: S51, mold size unit according to the mold size calculation formula "Dm1=(D M +Im+Es)*(1+Za),Dm2=D M +Im+Es+Bh+Cg,Dm3=D M3 *(1+So)-Cv, Dm1, Dm2, Dm3 are mold shape, structure, cavity radial size, D M is the size of the injection molded product, D M3 is the radial limit size of the injection molded part, Im is the inner margin of the mold, Es is the shrinkage at both ends of the injection molded part, Za is the mold processing allowance, Bh is the mold boss height, Cg is the mold gap, So is the shrinkage rate of the plastic material, and Cv is the mold correction value. The injection mold size data is obtained for the subsequent injection mold design number and passed to the molding pressure; S52, molding pressure unit according to the injection molding machine molding pressure calculation formula "Pi = P * Sa / [π * (d / 2) 2], P F =Sm*Pv / 1000, Pi is the injection molding machine injection pressure Pi (kg / cm 2 ), P is the injection molding machine pump pressure (kg / cm 2 ), Sa is the effective area of ​​the injection molding machine injection cylinder (cm 2 ), d is the screw diameter of the injection molding machine (cm), P F is the clamping force of the injection molding machine (T), Sm is the projected area of ​​the injection molding machine cavity (cm 2 ), Pv is the filling pressure of the injection molding machine (kg / cm 2 )” to obtain the molding injection pressure and clamping pressure respectively for subsequent mold molding debugging, and pass them to the injection weight unit; S53, shot weight unit is calculated according to the injection molding machine shot weight calculation formula "Wv = (π*(d / 2)2*Lt)*η*δ, Wv is the injection molding machine shot weight (g), d is the injection molding machine screw diameter (cm), Lt is the injection molding machine shot stroke (cm), η is the specific gravity of the plastic material (g / cm 3 ),δ is the injection molding machine efficiency” to obtain the injection molding weight of the injection molding machine for subsequent injection mold molding debugging and transfer to the ejector strength unit; S54, ejector strength unit is based on the injection mold ejector strength calculation formula "F = π2*λ*(π*(d o / 2)2)*E*r 2 / L2, F is the mold ejector pin buckling load (kgf), λ is the mold ejector pin support condition constant, d o is the diameter of the mold ejector pin (cm), E is the longitudinal elastic modulus of the mold ejector pin (kgf / cm 2 ), r is the inertia radius of the mold ejector section (cm), L is the length of the mold ejector (cm)” to obtain the ejector strength of the injection mold for subsequent injection mold molding debugging and pass it to the mold production unit; S55, the mold making unit designs and draws the mold according to the predetermined mold making process parameters, makes the mold through the mold making equipment, and obtains the mold processing quality information to ensure the quality of the subsequent molding trial, and transmits it to the molding debugging unit; S56, the molding debugging unit performs molding debugging through the injection molding machine according to the predetermined molding process parameters and obtains the quality information of the injection molded parts to ensure the accuracy of the mold design and transmits it to the processing center; S57, the processing center compares the mold quality information with the injection mold quality standard stored in the memory: if it meets the standard, it is set as the official injection mold process parameter; if it does not meet the standard, it is transmitted to the alarm and notified to adjust the parameters for rework. S58. The abnormality recognition unit matches the injection mold quality inspection image acquired in real time with the corresponding image in the trained convolutional neural network model and confirms its abnormality category for subsequent molding improvement and parameter adjustment.

4. The AI-based rapid injection mold design method according to claim 1 is characterized in that: The step S30 comprises the following steps: S31, input the filling unit and calculate the filling size according to the mold data "Ph=[(Ho-1)*Sh+Kh-Hi] / 2,Pw=[(Wo-1)*Sw+Kw-Wi] / 2,Ph,Pw are the padding sizes in height / width direction respectively,Hi,Wi are the height / width of input feature map respectively,Kh,Kw are the height / width of convolution kernel respectively,Sh,Sw are the values ​​of step length in height / width direction respectively,Ho,Wo are the height and width of output feature map respectively" obtain padding data and input, and pass it to convolution input unit; S32, the convolution input unit obtains the convolution input value and activates the function input according to the convolution input calculation formula "z(t) = ∫x(m)y(tm)dm, z(t) is the output function, x(t) is the input function, y(t) is the convolution kernel function, and dm is the integral differential element of the variable m", and passes it to the pooling conversion unit; S33, the pooling conversion unit obtains the maximum pixel value after pooling according to the maximum pooling calculation formula "Z(I,j)=max(X[i*Ps(i+1)*Ps,j*Ps(j+1)*Ps]", Z(I,j) is a pixel value in the output feature map after pooling, i and j are the positions in the output feature map, max is the maximum value of all pixel values ​​in the input window, X is the input feature map, and Ps is the window size of the pooling operation" to enter the fully connected layer to retain important feature information and pass it to the fully connected layer unit; S34, the fully connected layer unit obtains the output data of the connection layer according to the fully connected layer calculation formula "y=f(∑(Wn*Xn)+b), y is the output result, f is the activation function, Wn is the weight of the nth input feature, Xn is the nth input feature, n is the dimension of the input feature, and b is the bias" and enters the output layer for subsequent output calculation and is passed to the normalized output unit; S35, the normalized output unit obtains the output of Batch Norm according to the Batch Norm layer calculation formula "h = φ (Bn (Wx + b)), h is the output of the fully connected layer, φ is the activation function, Bn is the batch normalization operator, W is the weight parameter, X is the input of the fully connected layer, and b is the bias parameter", and passes it to the output conversion unit; S36, the output conversion unit obtains the output value size after convolution according to the output layer conversion calculation formula "N = (P-F + 2C) / S + 1, N is the output size after convolution, P is the input size before convolution, F is the convolution kernel size, C is the number of layers of 0 added around the image, and S is the step size" for subsequent model analysis and training, and passes it to the back propagation unit; S37, the back propagation unit obtains the loss function value and back propagates according to the loss function calculation formula "L = max(0, m + y*(f(x) - b)", L is the loss function value, m is the hyperparameter, y is the sample label (0 or 1), f(x) is the predicted value of the model, and b is the classification boundary" to optimize the performance of the network model; The step S60 comprises the following steps: S61, shrinkage confirmation unit calculates the shrinkage rate of injection molded parts according to the formula "Sr = [-D M +sqrt(4D M *Dm-3D M 2 )] / 2D M , Sr is the shrinkage rate, Dm is the mold size, D M Obtain the shrinkage rate of the injection molded parts for the "injection molded parts size" to confirm the accuracy of the injection molded parts and pass it to the appearance inspection unit; S62, the drawing determination unit improves the injection mold design drawing and its corresponding design information according to the parameters after the trial mold debugging to determine the final mold design plan, and transmits it to the appearance inspection unit; S63, the appearance inspection unit obtains the image or video of the injection mold through a high-definition camera and converts it into recognizable appearance quality information for subsequent packaging and shipment and molding mass production, and transmits it to the processing center; S64, the processing center compares the quality information of the mold with the quality standard of the injection mold stored in the memory: if it meets the standard, the customer is notified that mass production can be started; if it does not meet the standard, the alarm is transmitted to the alarm and rework is notified.

5. The AI-based rapid injection mold design method according to claims 1-4 is characterized in that: The system also includes a computer-readable storage medium containing a memory; the memory stores a computer program, and when the functional modules execute the computer program, the steps of the AI-based rapid injection mold design method described in any one of claims 1 to 4 are implemented; the computer-readable storage medium stores a computer program, and when the computer program is executed by the functional modules, the steps of the AI-based rapid injection mold design method described in any one of claims 1 to 4 are implemented; and also includes an AI-based injection mold design control device manufacturing control device, which is implemented by the AI-based rapid injection mold design method described in any one of claims 1 to 4.

6. An AI-based injection mold design control device, characterized in that: This is achieved by using an AI-based rapid injection mold design method as described in any one of claims 1 to 4 above.

Images