Injection system control method and device

Abstract

The invention discloses an injection system control method and device, and belongs to the technical field of die casting machine control, and the method comprises the steps: building a first model by taking a time sequence process parameter as input and taking a future time period equipment parameter change trend as output, and building a second model by taking the future time period equipment parameter change trend as input and taking a control parameter as output; dividing the injection process into a plurality of injection stages according to working states, training the first model and the second model according to historical process parameters, historical equipment parameters and historical control parameters of each injection stage, and combining to obtain a third model; and the obtained real-time process parameters are input into the third model, the third model is adjusted and optimized according to the real-time equipment parameters, target control parameters are obtained, and the parameters adjust the real-time control parameters of the injection system and / or switch the injection stage. The injection system is subjected to prediction and self-adaptive adjustment according to different injection stages, and switching of the injection stages is controlled, so that accurate control over the injection system is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of die-casting machine control, and in particular to a method and device for controlling an injection system. Background Art

[0002] Die casting is an efficient forming method widely used in the manufacture of non-ferrous metal products, especially in the automotive, communications, electronics and other industries. The die-casting machine injection process is usually divided into three stages: slow injection, fast injection and pressurization and pressure holding. Each stage has different requirements for process parameters such as injection speed and injection pressure. The control of the die-casting machine injection system directly affects the filling process of the molten metal, which in turn determines the dimensional accuracy, surface and internal quality of the final die-cast product. However, in the actual production process, due to changes in material types, temperature fluctuations, mold conditions and other process parameters, how to ensure the dynamic response and control accuracy of the injection system under complex working conditions has become a major challenge in the current field of die-casting machine control systems.

[0003] Traditional injection control systems mainly use PID control, which adjusts the injection speed and pressure by manually setting the proportional, integral and differential parameters (Kp, Ki and Kd). However, traditional PID controllers have response lag and overshoot problems when dealing with changes in working conditions, lack of adaptability, and strong manual dependence. In order to solve the limitations of traditional PID control, some intelligent control methods have emerged in recent years, such as fuzzy control and genetic algorithm optimization. These methods can improve the responsiveness and adaptability of the system to a certain extent, but their application in the injection control of die-casting machines still faces the following problems: First, the predictive ability is limited. Most of the existing intelligent control methods rely on real-time data feedback of the current working conditions, lack the ability to predict the future process status, and cannot predict the possible process changes in the system in advance, which easily leads to control lag; Second, the real-time and adaptability are insufficient. Although fuzzy control methods can optimize certain process parameters, they often find it difficult to adjust all relevant process parameters in real time and accurately during the die-casting process, especially in the rapidly changing injection stage; Third, there is a lack of dynamic adjustment mechanism. The existing intelligent control methods usually lack a dynamic adjustment mechanism for PID controller parameters, especially when the injection requirements of different stages (such as slow, fast, pressurization and pressure holding) in the die-casting process are significantly different, it is difficult to achieve stage-by-stage optimization.

[0004] Therefore, there is an urgent need for a method and device for controlling an injection system to solve the technical problem in the prior art that the injection parameters cannot be accurately controlled. Summary of the invention

[0005] In view of this, it is necessary to provide a method and device for controlling an injection molding system, which can predict the injection molding system in stages according to different injection molding stages and adaptively adjust the control parameters of the injection molding system, control the switching of the injection molding stages, so as to achieve precise control of the injection molding system.

[0006] In order to solve the above technical problems, on the one hand, the present invention provides a method for controlling an injection system, comprising: A first model is constructed with the timing process parameters as input and the equipment parameter change trend in the future time period as output, and a second model is constructed with the equipment parameter change trend in the future time period as input and the control parameters as output; The injection molding process is divided into a plurality of injection molding stages according to the working state, and the first model and the second model are trained according to the historical process parameters, the historical equipment parameters and the historical control parameters of each injection molding stage, and then the third model is obtained by merging them; The acquired real-time process parameters are input into the third model, and the third model is adjusted and optimized according to the real-time equipment parameters to obtain target control parameters, and the real-time control parameters of the injection system are adjusted and / or the injection stage is switched according to the target control parameters.

[0007] In a possible implementation, a first model is constructed with sequential process parameters as input and equipment parameter change trends in future time periods as output, including: Based on the LSTM neural network, the first model is constructed with the timing process parameters as input and the change trend of the equipment parameters in the future time period as output, wherein the first model includes a first input layer, multiple LSTM layers, a fully connected layer, and a first output layer; The input layer is used to receive historical timing process parameters; The LSTM layer is used to process historical time series process parameters, determine the dependencies of time series in the time series process parameters, and extract time series features according to the dependencies; The fully connected layer is used to connect the LSTM layer and the output layer, perform feature mapping on the time series features, and obtain mapping features; The output layer is used to output the change trend of the equipment parameters in the future time period according to the mapping features.

[0008] In a possible implementation, a second model is constructed with the equipment parameter change trend as input and the control parameter as output, including: Based on the RBF neural network, a second model is constructed with the equipment parameter change trend as input and the control parameter as output, wherein the second initial prediction model includes a second input layer, multiple hidden layers and a second output layer; The second input layer is used to receive the device parameter change trend; The hidden layer is used to extract characteristic information of equipment parameter change trends; The second output layer is used to output control parameters according to the feature information.

[0009] In a possible implementation, the first model and the second model are trained according to the historical process parameters, equipment parameters and control parameters of each injection stage, and then combined to obtain a third model, including: Obtain historical timing process parameters and historical timing equipment parameters at different injection stages; Iteratively training the first model according to the historical timing process parameters and the historical timing equipment parameters to obtain a first target model; Obtain historical equipment parameters and historical control parameters at different injection stages; Iteratively training the second model according to the historical equipment parameters and the historical control parameters to obtain a second target model; The first target model and the second target model constitute a third model.

[0010] In a possible implementation, iteratively training the first model according to the historical timing process parameters and the historical timing equipment parameters to obtain a first target model includes: Based on the LSTM layer, the historical time series process parameters are forward propagated to calculate the hidden state and memory unit state of each time step; Generate a prediction value through a fully connected layer according to the hidden state and the memory unit state, and output a device parameter change trend in a future time period through a first output layer; Optimizing the first model according to a preset loss function, historical time series predicted equipment parameters, and equipment parameter change trends in future time periods to obtain a first optimized model; Based on the optimization algorithm, the output weights and biases of the first optimization model are adjusted according to the gradient of the time step to obtain the first target model.

[0011] In a possible implementation, the optimization algorithm is the Adam optimization algorithm, and the calculation formula of the output weight is: , in, is the weight at time t, and are respectively the momentum and second-order moment estimates of the gradient, is the learning rate, is a small constant that prevents the denominator from being zero.

[0012] In a possible implementation, iteratively training the second model according to the historical device parameters and the historical control parameters to obtain a second target model includes: Cluster the historical equipment parameters based on the K-means clustering algorithm to obtain multiple cluster center points; Activating the hidden layer based on a Gaussian function, and calculating a width parameter of the cluster center point; Training the weight of the second output layer according to the width parameter, and outputting the control parameter prediction result according to the weight; Based on the preset loss function and control parameter prediction results, the second model is iteratively trained to obtain a second target model.

[0013] In a possible implementation, the third model is adjusted and optimized according to the real-time device parameters to obtain the target control parameters, including: Obtain real-time equipment parameters during the die-casting machine injection process; Adaptively optimizing and adjusting the output result of the first model in the third model according to the real-time device parameters; The output result of the first model after optimization and adjustment is used as the input of the second model to obtain the final target control parameters.

[0014] In a possible implementation, the injection stage includes a slow injection stage, a fast injection stage and a pressure-increasing and pressure-holding stage, and the injection stage of switching the injection system according to the real-time control parameter includes: When the real-time control parameter is in the preset slow-speed injection interval, the injection system is controlled to be in the slow-speed injection stage; When the real-time control parameter is in the preset fast-injection interval, the injection system is controlled to be in the fast-injection stage; When the real-time control parameter is in the preset pressure-increasing and pressure-maintaining interval, the injection system is controlled to be in the pressure-increasing and pressure-maintaining stage; When the real-time control parameter reaches the preset upper limit of the pressure increase and pressure maintenance, the injection system ends its operation.

[0015] In a second aspect, the present invention further provides an injection system control device, comprising: A model building module, for building a first model with a time sequence process parameter as input and a change trend of equipment parameters in a future time period as output, and building a second model with a change trend of equipment parameters in a future time period as input and a control parameter as output; A model training module, used for dividing the injection molding process into multiple injection molding stages according to the working state, training the first model and the second model according to the historical process parameters, equipment parameters and control parameters of each injection molding stage, and merging them to obtain a third model; The injection system control module is used to adjust and optimize the third model according to the real-time process parameters and the real-time equipment parameters to obtain target control parameters, adjust the real-time control parameters according to the target control parameters and / or switch the injection stage.

[0016] The beneficial effects of the present invention are as follows: first, based on the LSTM neural network, a first time series model is constructed with time series process parameters as input and equipment parameter change trends in future time periods as output, and the equipment parameter change trends in future time periods can be predicted based on historical time series process parameters and current process parameters; based on the RBF neural network, the equipment parameter change trends in future time periods are input and control parameters are output, and a second model is constructed to further predict the control parameters of the PID based on the equipment parameters predicted by the LSTM; then, the injection process is divided into a plurality of injection stages according to the working state of the injection system, and the first model and the second model are iteratively trained according to the historical process parameters, equipment parameters and control parameters of each injection stage to obtain a third model for predicting control parameters, and the model is trained in stages according to the parameter characteristics of different injection stages, so that the model can accurately predict the parameters of each stage; finally, according to the real-time process parameters and the trained third model, according to the real-time equipment parameters monitored in real time, the output results of the first model in the third model are adaptively adjusted and optimized, and the target control parameters are finally obtained, the real-time control parameters of the PID are adjusted according to the target control parameters, and the injection system is switched to different injection stages according to the adjusted real-time control parameters. The present invention constructs an LSTM-RBF fusion model, divides the injection process of the injection system into different injection stages, predicts the control parameters of the PID controller according to real-time process parameters, and switches the working stage of the injection system according to the control parameters. It realizes accurate prediction of the control parameters of the PID controller and adaptive adjustment of the control parameters and injection stages of the injection system, thereby improving the control accuracy of the injection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A schematic flow chart of an embodiment of the injection system control method provided by the present invention; Figure 2 The present invention provides Figure 1 A schematic diagram of a flow chart of an embodiment of step S102; Figure 3 The present invention provides Figure 2 A schematic diagram of a flow chart of an embodiment of step S202; Figure 4 The present invention provides Figure 2Schematic diagram of the process of the first embodiment of step S204; Figure 5 A schematic diagram of the switching process of each injection stage provided by the present invention; Figure 6 A schematic structural diagram of an embodiment of the injection system control device provided by the present invention. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0020] In the description of the embodiments of the present invention, unless otherwise specified, "multiple" means two or more than two. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" may mean: A exists alone, A and B exist at the same time, and B exists alone.

[0021] The descriptions of "first", "second", etc. involved in the embodiments of the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the technical features defined as "first" or "second" may explicitly or implicitly include at least one of the features.

[0022] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0023] Before describing the embodiments, the following definitions are given for the relevant terms: LSTM neural network is a time-recurrent neural network that aims to solve the gradient vanishing or exploding problem faced by traditional RNN when processing long sequence data, so that it can capture long-term dependencies more effectively. The RFB network is a network structure for target detection, which aims to enhance the feature extraction capability of the network by simulating the receptive field of human vision.

[0024] The present invention provides a method and device for controlling an injection system, which are described below respectively.

[0025] Figure 1 A flow chart of an embodiment of the injection system control method provided by the present invention is as follows: Figure 1 As shown, the injection system control method includes: S101, constructing a first model with the timing process parameters as input and the equipment parameter change trend in the future time period as output, and constructing a second model with the equipment parameter change trend in the future time period as input and the control parameters as output; It should be noted that a time series model and a target detection model are constructed through the model management software on the electronic terminal. In this embodiment, an LSTM neural network is used to construct a first time series model, and an RBF neural network is used to construct a target detection model, but it is not limited to the LSTM neural network and the RBF neural network. The electronic terminal can realize data processing and model management functions, and can be a PC, mobile terminal, portable terminal, etc.

[0026] It should be further explained that the timing process parameters refer to a series of key parameters that need to be set and adjusted during the die-casting process, which determine the quality and production efficiency of the die-casting parts, including pressure parameters (injection pressure, pressure holding time, etc.), temperature parameters (melting temperature, mold temperature, casting cooling temperature, etc.), time parameters (injection time, cooling time, etc.) and speed parameters (injection speed, mold opening and closing speed, etc.) in different injection stages; pressure parameters affect the filling and solidification process of the molten metal, temperature parameters affect the molding and cooling of the casting, time parameters determine the time required for each stage in the die-casting process, and speed parameters affect the filling speed of the molten metal and the opening and closing efficiency of the mold; equipment parameters are the working status parameters of the die-casting machine and its ancillary equipment, such as the displacement of the valve core of the servo cartridge valve of the injection accumulator, the displacement of the valve core of the servo cartridge valve of the boost accumulator, and the holding time; control parameters are the parameters required in the control algorithm of the PID controller. The set parameters determine the response speed and stability of the control algorithm to the die-casting process, including the proportional coefficient, integral coefficient and differential coefficient. Among them, the proportional coefficient determines the response speed of the control algorithm to the deviation. The larger the proportional coefficient, the faster the response speed, but too large a proportional coefficient will cause overshoot and oscillation. The integral coefficient determines the response of the control algorithm to the cumulative effect of the deviation. The larger the integral coefficient, the more sensitive the system is to the cumulative effect of the deviation, which helps to eliminate static errors, but too large an integral coefficient will cause integral saturation and response lag. The differential coefficient determines the response of the control algorithm to the rate of change of the deviation. The larger the differential coefficient, the more sensitive the system is to the rate of change of the deviation, which helps to predict and suppress the change of the deviation in advance, but too large a differential coefficient will lead to sensitivity to noise. In the injection process of the die-casting machine, by reasonably setting and adjusting these parameters, precise control of the injection system can be achieved and production effects can be optimized.

[0027] S102, dividing the injection molding process into multiple injection molding stages according to the working state, training the first model and the second model according to the historical process parameters, historical equipment parameters and historical control parameters of each injection molding stage, and merging them to obtain a third model; It should be noted that the working status of the die-casting machine, such as the injection speed and injection pressure, and its control parameters also change with time. There are large differences between the process parameters, equipment parameters and control parameters in different stages. Dividing the injection process of the entire die-casting machine according to its working status and analyzing and processing the parameters of each stage can more accurately predict the control parameters of the PID controller.

[0028] S103, inputting the acquired real-time process parameters into the third model, and adjusting and optimizing the third model according to the real-time equipment parameters to obtain target control parameters, and adjusting the real-time control parameters of the injection system and / or switching the injection stage according to the target control parameters.

[0029] It should be noted that, in the present embodiment, the injection displacement sensor, injection pressure sensor, cartridge valve position sensor, mold temperature sensor, mold cavity pressure sensor, etc. are arranged at key positions of the die-casting machine to collect key process parameters and equipment parameters related to the injection process in real time, and the equipment parameter change trend in the future time period is obtained according to the real-time process parameters, the real-time equipment parameters are fed back to the first model, the equipment parameter change trend of the output of the first model is adjusted and optimized, and the target control parameters are obtained according to the optimized equipment parameter change trend in the future time period.

[0030] This embodiment predicts the control parameters of the PID controller by constructing an LSTM model and an RBF model; then, the injection process is divided into multiple injection stages according to the working state of the injection system, and the first model and the second model are iteratively trained according to the historical process parameters, equipment parameters and control parameters of each injection stage to obtain a third model for predicting the control parameters, and the model is trained in stages according to the parameter characteristics of different injection stages, so that the model can accurately predict the parameters of each stage; finally, according to the third model trained with real-time process parameters, according to the real-time equipment parameters monitored in real time, the output results of the first model in the third model are adaptively adjusted and optimized, and finally, the target control parameters are obtained, the real-time control parameters of the PID are adjusted according to the target control parameters, and the injection system is switched to different injection stages according to the adjusted real-time control parameters. This embodiment constructs an LSTM-RBF fusion model, divides the injection process of the injection system into different injection stages, predicts the control parameters of the PID controller according to the real-time process parameters, and switches the working stage of the injection system according to the control parameters, so as to achieve accurate prediction of the control parameters of the PID controller and adaptive adjustment of the control parameters and injection stages of the injection system, thereby improving the accuracy of injection system control.

[0031] In some embodiments of the present invention, a first model is constructed with sequential process parameters as input and sequential device parameters as output, including: Based on the LSTM neural network, the first model is constructed with the time series process parameters as input and the change trend of the equipment parameters in the future time period as output. The first initial prediction model includes a first input layer, multiple LSTM layers, a fully connected layer, and a first output layer. The input layer is used to receive historical timing process parameters; It should be noted that the received timing process parameters are pre-processed, including but not limited to using wavelet denoising technology to eliminate noise interference to improve data quality.

[0032] The LSTM layer is used to process historical time series process parameters, determine the dependencies of the time series in the time series process parameters, and extract time series features based on the dependencies; It should be noted that the LSTM layer is the core part of the LSTM neural network, which is composed of multiple LSTM units. Each LSTM unit includes four key components: forget gate, input gate, candidate memory unit, and output gate. The multi-layer cascade structure (such as 2-3 layers of LSTM) enhances the ability to capture complex temporal dependencies.

[0033] The fully connected layer is used to connect the LSTM layer and the output layer, perform feature mapping on the time series features, and obtain mapping features; The output layer is used to output the changing trend of equipment parameters in the future time period according to the mapping features.

[0034] It should be noted that the first model analyzes and learns the relationship between historical time-series process parameters and historical equipment parameters, and predicts the equipment parameter change trend in the future based on real-time process parameters.

[0035] In this embodiment, the LSTM neural network model can effectively preserve the important features in the time series, analyze and learn the process parameters and equipment parameters at different time points, and predict the trend of equipment parameter changes in the future, providing an accurate basis for subsequent data analysis.

[0036] In some embodiments of the present invention, a first model is constructed with sequential process parameters as input and sequential device parameters as output, including: The second model is constructed with the device parameters as input and the control parameters as output, including: Based on the RBF neural network, the second model is constructed with the equipment parameter change trend as input and the control parameter as output, and the second initial prediction model includes a second input layer, multiple hidden layers and a second output layer; The second input layer is used to receive the changing trend of equipment parameters; Specifically, the output result of the first model, the equipment parameter change trend in the future time period, is extracted through sliding window statistics (such as mean, variance, gradient) or time series analysis (such as difference, moving average, etc.), and these dynamic characteristics and the equipment parameters collected in real time are normalized.

[0037] The hidden layer is used to extract characteristic information of equipment parameter change trends; The second output layer is used to output control parameters according to the feature information.

[0038] It should be noted that the second output layer performs weighted summation on the outputs of the hidden layer to obtain the control parameters of the PID controller, namely, the proportional coefficient, the integral coefficient and the differential coefficient.

[0039] Specifically, the calculation formula of the second output layer is: , in, It is The weights from the hidden layer to the output, is the bias term and N is the number of hidden layers.

[0040] This embodiment uses the RBF neural network to analyze and predict the change trend of equipment parameters in the future time period, determine the control parameters of the PID controller, and improve the accuracy of control parameter prediction.

[0041] In some embodiments of the present invention, Figure 2 As shown, Figure 2 The present invention provides Figure 1 The flowchart of an embodiment of step S102 in the embodiment includes: S201, obtaining historical timing process parameters and historical timing equipment parameters of different injection stages; It should be noted that the process parameters and equipment parameters during the injection process are collected by various sensors arranged at key positions of the die-casting machine.

[0042] S202, iteratively training the first model according to historical timing process parameters and historical timing equipment parameters to obtain a first target model; The historical time series process parameters are denoised and normalized, and the learning ability and prediction accuracy of the network are optimized by setting the appropriate number of layers and the number of units in each layer. Then, the LSTM neural network is trained and learned on these historical data to capture the time series dependencies in the data and predict the changing trends of equipment parameters in the injection process at future moments, such as the displacement of the valve core of the servo cartridge valve of the injection accumulator, the displacement of the valve core of the servo cartridge valve of the booster accumulator, and the pressure holding time requirement in the future time step.

[0043] S203, obtaining historical equipment parameters and historical control parameters at different injection stages; It should be noted that various sensors are arranged at key positions of the die-casting machine to collect equipment parameters during the injection process, extract control data from the historical operation of the die-casting machine, and use these historical data as training data.

[0044] S204: Iteratively train the second model according to historical device parameters and historical control parameters to obtain a second target model.

[0045] It should be noted that the RBF network is trained and learned through the historical data of the die-casting machine. The data input in the training and learning includes historical key equipment parameters (such as the valve core displacement of the servo cartridge valve of the injection accumulator, the valve core displacement of the servo cartridge valve of the boost accumulator, and the holding time) and historical PID controller control parameters.

[0046] S205. The first target model and the second target model form a third model.

[0047] It should be noted that the trained LSTM first target model and RBF second target model are combined to obtain the LSTM-RBF third model for predicting control parameters. At the same time, during the long-term operation of the injection system, the third model is continuously optimized, and the LSTM and RBF network models are periodically updated using an iterative learning strategy. The LSTM and RBF networks update the model by expanding historical data and re-optimizing network weights; at the same time, according to the actual operating status of the system, the hyperparameters of the LSTM and RBF neural networks (such as learning rate, number of hidden layers, etc.) are further optimized to improve the system's adaptability and control accuracy. The frequency of iterative learning is set according to actual production needs. The system updates the model at regular intervals or when large fluctuations in the injection process parameters are detected to ensure the system's dynamic adaptive control capabilities under complex working conditions.

[0048] In some embodiments of the present invention, Figure 3 As shown, Figure 3 The present invention provides Figure 2 The flowchart of an embodiment of step S202 in the embodiment includes: S301, based on the LSTM layer, forward propagate the historical time series process parameters and calculate the hidden state and memory unit state of each time step; Specifically, during the training process, the LSTM network will forward propagate the input historical data of the die-casting machine (such as mold temperature, injection speed, and injection pressure), and perform forward propagation calculations through each unit of the LSTM layer to obtain the hidden state and memory unit state of each time step.

[0049] S302, generating a prediction value through a fully connected layer according to the hidden state and the memory unit state, and outputting the device parameter change trend in the future time period through the first output layer; S303, optimizing the first model according to a preset loss function, historical time series prediction of device parameters, and device parameter change trends in future time periods to obtain a first optimized model; Specifically, the mean square error loss function (MSE) is used to calculate the difference between the predicted value and the input historical key equipment parameters: , in, is the loss value, N is the number of samples, and are the true value and predicted value of the tth sample respectively.

[0050] S304: Based on the optimization algorithm, the output weight and bias of the first optimization model are adjusted according to the gradient of the time step to obtain a first target model.

[0051] Specifically, by back-propagating to each time step, the gradient of each time step is calculated during the back-propagation process, and the gradient is calculated by forward propagation step by step. This embodiment does not specifically limit the optimization algorithm, and preferably uses the Adam optimization algorithm to update the weights and biases of the network. The calculation formula of the dynamic weight is: , in, is the weight at time t, and are respectively the momentum and second-order moment estimates of the gradient, is the learning rate, is a small constant that prevents the denominator from being zero.

[0052] Furthermore, during the iterative training process, if the loss function remains unchanged or becomes very small in multiple iterations, the model is deemed to have converged and the training process is terminated; and the training effect is tested through a test set, the timing data in the test set (such as mold temperature, injection speed, injection pressure) is input, and the weights and biases learned during training are used to perform forward propagation to generate prediction results; if the error between the prediction result and the historical key equipment parameters in the test set is within 1%, the training result is determined to be satisfactory; if the error is too large, the model parameters (such as learning rate, number of LSTM layers, number of hidden layers) are adjusted or the number of training rounds is increased.

[0053] This embodiment optimizes the performance of the model through iterative training of the LSTM model, thereby improving the accuracy of control parameter prediction.

[0054] In some embodiments of the present invention, Figure 2 As shown, Figure 2 The present invention provides Figure 1 The flowchart of an embodiment of step S204 in the embodiment includes: S401, clustering historical equipment parameters based on K-means clustering algorithm to obtain multiple cluster center points; Specifically, multiple cluster center points are randomly initialized, the Euclidean distance between each device parameter and the center point is calculated, the data is assigned to the nearest cluster, the cluster center is updated to the mean value of the data within the cluster, and it is iterated until the center point is stable to obtain multiple cluster center points; S402, activating the hidden layer based on the Gaussian function, and calculating the width parameter of the cluster center point; Specifically, based on the Gaussian function, the width parameter of each center point is calculated as follows: , in, is the input vector, It is The center vector of the cluster center points, It is The width parameter of the cluster centers, is a Gaussian function.

[0055] S403, training the weight of the second output layer according to the width parameter, and outputting the control parameter prediction result according to the weight; Specifically, the output layer maps the output of the hidden layer to the target control parameters, namely the proportional, integral, and differential parameters of the PID controller; the output layer trains the weights by minimizing the objective function, and its formula is: , Among them, c is the control parameter, is the target output vector, i.e., the control parameters of the PID controller, is the weight matrix of the output layer, is the hidden layer output matrix, and this process is solved by the least squares method.

[0056] S404: Based on the preset loss function and control parameter prediction results, the second model is iteratively trained to obtain a second target model.

[0057] Specifically, the mean square error loss function is used to minimize the error between the output layer prediction value and the true value: , Where N is the number of samples, and are the true output of the ith sample and the model prediction value, respectively. When the loss function value does not decrease significantly in multiple consecutive training rounds, the model is considered to have converged and the training process is completed. The training effect is tested through the test set, and the key equipment parameters in the test set (such as the displacement of the valve core of the servo cartridge valve of the injection accumulator, the displacement of the valve core of the servo cartridge valve of the booster accumulator, and the pressure holding time) are input. The weights learned during training are used to generate the prediction results. If the error between the prediction result and the PID controller control parameter in the test set is within 1%, the training result is considered to be satisfactory. If the error is too large, the model parameters (the center vector of the RBF neuron) are adjusted. , the width parameter of RBF neurons , number of hidden layer neurons, etc.).

[0058] In some embodiments of the present invention, the injection stage includes a slow injection stage, a fast injection stage and a pressure-increasing and pressure-holding stage; the injection stage of switching the injection system according to the real-time control parameters includes: When the real-time control parameters are in the preset slow-speed injection range, the injection system is controlled to be in the slow-speed injection stage; When the real-time control parameters are in the preset fast-injection interval, the injection system is controlled to be in the fast-injection stage; When the real-time control parameters are in the preset pressure-increasing and pressure-holding range, the injection system is controlled to be in the pressure-increasing and pressure-holding stage; When the real-time control parameter reaches the preset upper limit of boost pressure and pressure maintenance, the injection system ends its operation.

[0059] It should be noted that in the injection process of the die-casting machine, the injection speed is low in the initial stage to avoid splashing of the molten metal and reduce the entrapment of gas. Subsequently, the molten metal is filled at a high speed to ensure complete molding and reduce residual gas in the cavity. Finally, after the cavity is filled, the pressure is increased rapidly to solidify the molten metal under high pressure and increase the density of the casting until the pressure reaches the preset threshold and the holding time reaches the preset threshold, ending the injection process. Therefore, the injection process of the die-casting machine is divided into a slow injection stage, a fast injection stage, and a pressurization and holding stage.

[0060] Specifically, the position of the injection punch is controlled according to the PID controller control parameters output by the third model, such as Figure 5 As shown, Figure 5 The present invention provides a schematic diagram of the switching process of each injection stage. When the position of the injection punch is in the preset slow injection interval, the injection system is ensured to be in the slow injection stage; when the position of the injection punch reaches the preset fast injection interval, the injection system is switched to the fast injection stage; when the position of the injection punch reaches the preset boost and pressure holding interval, the injection system is switched to the boost and pressure holding stage; when the pressure and the pressure holding time both reach the preset upper limit values, the injection work is ended.

[0061] In order to better implement the injection system control method in the embodiment of the present invention, based on the injection system control method, correspondingly, Figure 6 As shown, the embodiment of the present invention further provides an injection system control device 600 including: Model building module 601, used to build a first model with the timing process parameters as input and the equipment parameter change trend in the future time period as output, and to build a second model with the equipment parameter change trend in the future time period as input and the control parameters as output; A model training module 602 is used to divide the injection process into multiple injection stages according to the working state, train the first model and the second model according to the historical process parameters, equipment parameters and control parameters of each injection stage, and then combine them to obtain a third model; The injection system control module 603 is used to adjust and optimize the third model according to the real-time process parameters and the real-time equipment parameters to obtain target control parameters, adjust the real-time control parameters according to the target control parameters and / or switch the injection stage.

[0062] The injection system control device 600 provided in the above embodiment can implement the technical solution described in the above injection system control method embodiment. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above injection system control method embodiment, which will not be repeated here.

[0063] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0064] The above is a detailed introduction to the injection system control method and device provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for controlling an injection system, characterized in that: include: A first model is constructed with the timing process parameters as input and the equipment parameter change trend in the future time period as output, and a second model is constructed with the equipment parameter change trend in the future time period as input and the control parameters as output; The injection molding process is divided into multiple injection molding stages according to the working state, and the first model and the second model are trained according to the historical process parameters, historical equipment parameters and historical control parameters of each injection molding stage, and then the third model is obtained by merging them; The acquired real-time process parameters are input into the third model, and the third model is adjusted and optimized according to the real-time equipment parameters to obtain target control parameters, and the real-time control parameters of the injection system are adjusted and / or the injection stage is switched according to the target control parameters.

2. The method for controlling the injection system according to claim 1, characterized in that: The first model is constructed with the sequential process parameters as input and the equipment parameter change trend in the future time period as output, including: Based on the LSTM neural network, the first model is constructed with the timing process parameters as input and the change trend of the equipment parameters in the future time period as output, wherein the first model includes a first input layer, multiple LSTM layers, a fully connected layer, and a first output layer; The input layer is used to receive historical timing process parameters; The LSTM layer is used to process historical time series process parameters, determine the dependencies of time series in the time series process parameters, and extract time series features according to the dependencies; The fully connected layer is used to connect the LSTM layer and the output layer, perform feature mapping on the time series features, and obtain mapping features; The output layer is used to output the change trend of the equipment parameters in the future time period according to the mapping features.

3. The injection system control method according to claim 2, characterized in that: The second model is constructed with the equipment parameter change trend as input and the control parameter as output, including: Based on the RBF neural network, a second model is constructed with the equipment parameter change trend as input and the control parameter as output, wherein the second model includes a second input layer, a plurality of hidden layers and a second output layer; The second input layer is used to receive the device parameter change trend; The hidden layer is used to extract characteristic information of equipment parameter change trends; The second output layer is used to output control parameters according to the feature information.

4. The injection system control method according to claim 3, characterized in that: After training the first model and the second model according to the historical process parameters, equipment parameters and control parameters of each injection stage, a third model is obtained by merging, including: Obtain historical timing process parameters and historical timing equipment parameters at different injection stages; Iteratively training the first model according to the historical timing process parameters and the historical timing equipment parameters to obtain a first target model; Obtain historical equipment parameters and historical control parameters at different injection stages; Iteratively training the second model according to the historical equipment parameters and the historical control parameters to obtain a second target model; The first target model and the second target model constitute a third model.

5. The method for controlling an injection system according to claim 4, characterized in that: Iteratively training the first model according to the historical timing process parameters and the historical timing equipment parameters to obtain a first target model includes: Based on the LSTM layer, the historical time series process parameters are forward propagated to calculate the hidden state and memory unit state of each time step; Generate a prediction value through a fully connected layer according to the hidden state and the memory unit state, and output a device parameter change trend in a future time period through a first output layer; Optimizing the first model according to a preset loss function, historical time series predicted equipment parameters, and equipment parameter change trends in future time periods to obtain a first optimized model; Based on the optimization algorithm, the output weights and biases of the first optimization model are adjusted according to the gradient of the time step to obtain the first target model.

6. The method for controlling the injection system according to claim 5, characterized in that: The optimization algorithm is the Adam optimization algorithm, and the calculation formula of the output weight is: ， in, is the weight at time t, and are respectively the momentum and second-order moment estimates of the gradient, is the learning rate, is a small constant that prevents the denominator from being zero.

7. The method for controlling an injection system according to claim 4, wherein: Iteratively training the second model according to the historical equipment parameters and the historical control parameters to obtain a second target model includes: Cluster the historical equipment parameters based on the K-means clustering algorithm to obtain multiple cluster center points; Activating the hidden layer based on a Gaussian function, and calculating a width parameter of the cluster center point; Training the weight of the second output layer according to the width parameter, and outputting the control parameter prediction result according to the weight; Based on the preset loss function and control parameter prediction results, the second model is iteratively trained to obtain a second target model.

8. The method for controlling an injection system according to claim 1, wherein: The third model is adjusted and optimized according to the real-time device parameters to obtain target control parameters, including: Obtain real-time equipment parameters during the die-casting machine injection process; Adaptively optimizing and adjusting the output result of the first model in the third model according to the real-time device parameters; The output result of the first model after optimization and adjustment is used as the input of the second model to obtain the final target control parameters.

9. The method for controlling an injection system according to claim 1, wherein: The injection stage includes a slow injection stage, a fast injection stage and a pressure-increasing and pressure-maintaining stage. The injection stage of the injection system is switched according to the real-time control parameters, including: When the real-time control parameter is in the preset slow-speed injection interval, the injection system is controlled to be in the slow-speed injection stage; When the real-time control parameter is in the preset fast-injection interval, the injection system is controlled to be in the fast-injection stage; When the real-time control parameter is in the preset pressure-increasing and pressure-maintaining interval, the injection system is controlled to be in the pressure-increasing and pressure-maintaining stage; When the real-time control parameter reaches the preset upper limit of the pressure increase and pressure maintenance, the injection system ends its operation.

10. A control device for an injection system, characterized in that: include: A model building module, for building a first model with a time sequence process parameter as input and a change trend of equipment parameters in a future time period as output, and building a second model with a change trend of equipment parameters in a future time period as input and a control parameter as output; A model training module, used for dividing the injection molding process into multiple injection molding stages according to the working state, training the first model and the second model according to the historical process parameters, equipment parameters and control parameters of each injection molding stage, and merging them to obtain a third model; The injection system control module is used to adjust and optimize the third model according to the real-time process parameters and the real-time equipment parameters to obtain target control parameters, adjust the real-time control parameters according to the target control parameters and / or switch the injection stage.

Images