Die casting machine injection speed control method and system based on LSTM neural network

Abstract

The application provides a die casting machine injection speed control method and system based on an LSTM neural network to solve the problem of large injection speed linear model error and inaccurate data of the current die casting machine injection; the method comprises the following steps: A1: obtaining multiple injection data information of the die casting machine injection; A2: establishing an injection system LSTM neural network inversion model; A3: obtaining a preset injection valve core displacement corresponding to a preset injection speed of the die casting machine injection through the injection system LSTM neural network inversion model; A4: controlling the injection valve core to move the preset injection valve core displacement and controlling the corresponding die casting machine to perform injection. The method can accurately predict the valve core displacement corresponding to the preset injection speed by establishing the injection system LSTM neural network inversion model, and provides a more accurate initial injection speed and / or valve core displacement data input for subsequent iterative learning.

Description

TECHNICAL FIELD

[0001] The application relates to the field of injection speed control of a die casting machine, and in particular to an injection speed control method and system for a die casting machine based on an LSTM neural network. BACKGROUND

[0002] There are two common control methods for adjusting the injection speed of a die casting machine. The first method is to manually adjust the spool valve to adjust the injection speed. The second method is to use an improved electrically adjusted spool valve, that is, to replace the manually adjusted valve core displacement with a servo motor. The electrically adjusted spool valve cannot achieve real-time control of the injection speed, and can only adjust the valve core displacement after each injection to achieve the desired injection speed in the next injection. Both of the above-mentioned control methods for adjusting the injection speed require adjusting the corresponding valve core displacement to achieve the desired injection speed.

[0003] The current method for predicting the valve core displacement is to collect the injection speeds corresponding to a plurality of different valve core displacements, and to establish a simple linear model according to the linear relationship between the valve core displacement and the corresponding injection speed, so as to predict the valve core displacement under the desired injection speed.

[0004] However, the relationship between the injection speed of the die casting machine and the valve core displacement is highly nonlinear. The linear model of the speed system established by this method has a large error, and cannot accurately predict the valve core displacement under the desired injection speed, and cannot achieve the desired injection speed. SUMMARY

[0005] To solve the problem of large error of the existing linear model, and to effectively solve the problem of manually adjusting the valve core displacement by experienced operators to achieve the desired injection speed, the application provides an injection speed control method and system for a die casting machine based on an LSTM neural network. To solve the above-mentioned problems, the application adopts the following technical solutions:

[0006] The injection speed control method for a die casting machine based on an LSTM neural network comprises the following steps:

[0007] A1: acquiring a plurality of injection data information of the die casting machine, wherein the injection data information comprises the injection speed of the die casting machine and the corresponding injection valve core displacement;

[0008] A2: combining the acquired plurality of injection data information and a preset LSTM neural network model, establishing an injection system LSTM neural network inversion model according to a preset model algorithm;

[0009] A3: obtaining the preset injection valve core displacement corresponding to the preset injection speed of the die casting machine through the injection system LSTM neural network inversion model;

[0010] A4: controlling the injection valve core to move the preset injection valve core displacement and controlling the corresponding die casting machine to inject.

[0011] Further, the method further comprises the steps of:

[0012] A5: obtaining an actual injection speed of the corresponding die casting machine when the corresponding die casting machine is controlled to inject according to the preset injection valve core displacement;

[0013] A6: determining whether the actual injection speed is an injection speed within a preset error range;

[0014] A7: if yes, continuously maintaining the preset injection valve core displacement corresponding to the actual injection speed to control the corresponding die casting machine to inject;

[0015] A8: if no, performing iterative learning on the actual injection speed according to a preset iterative learning algorithm until the actual injection speed of the die casting machine is an injection speed within the preset error range.

[0016] Further, the step A6 of determining whether the actual injection speed is an injection speed within a preset error range comprises the steps of:

[0017] obtaining a preset actual speed difference between the preset injection speed and the actual injection speed;

[0018] determining whether the preset actual speed difference is within the preset error range;

[0019] if yes, the actual injection speed is an injection speed within the preset error range;

[0020] if no, the actual injection speed is not an injection speed within the preset error range.

[0021] Further, the step A8 comprises the steps of:

[0022] obtaining a verification injection valve core displacement corresponding to the actual injection speed through an injection system LSTM neural network inversion model;

[0023] obtaining a displacement difference between the preset injection valve core displacement and the verification injection valve core displacement;

[0024] inputting the obtained displacement difference into a preset proportional iterative learning algorithm, and obtaining an iterative injection valve core displacement;

[0025] controlling the injection valve core to move the iterative injection valve core displacement and controlling the corresponding die casting machine to inject;

[0026] obtaining an actual injection speed of the corresponding die casting machine when the corresponding die casting machine is controlled to inject according to the iterative injection valve core displacement;

[0027] Continuing to execute step A6 until the actual injection speed of the die casting machine is within the preset error range of the injection speed.

[0028] Further, the acquired multiple injection data information and the preset LSTM neural network model are input into the preset model establishment algorithm to establish an injection system LSTM neural network inversion model.

[0029] Acquire multiple injection speeds of the die casting machine and corresponding injection valve core displacement amounts;

[0030] Data preprocessing is performed on the injection speed of the die casting machine to acquire injection speed data information after data preprocessing;

[0031] The injection speed data information after data preprocessing and the corresponding injection valve core displacement amounts are input into the preset LSTM neural network model;

[0032] Iterative training is performed;

[0033] An injection system LSTM neural network inversion model is established.

[0034] Further, the data preprocessing includes the following steps:

[0035] Acquire injection speed data information in the injection data information of the die casting machine;

[0036] Through wavelet decomposition, the injection speed data information is decomposed according to preset high-frequency speed information and preset low-frequency speed information;

[0037] The preset high-frequency speed information is denoised through a wavelet threshold denoising algorithm;

[0038] The denoised data information is reconstructed according to a preset reconstruction algorithm;

[0039] Acquire the reconstructed injection speed curve information.

[0040] Further, in step A8, the preset proportional iterative learning algorithm is:

[0041] X f2 =X f1 +K△X

[0042] Wherein X f2 is: the iterative injection valve core displacement amount;

[0043] Wherein X f1 is: the preset injection valve core displacement amount;

[0044] Wherein △X is: the displacement amount difference between the preset injection valve core displacement amount and the verification injection valve core displacement amount;

[0045] K is a proportional coefficient.

[0046] Die casting machine injection speed control system based on LSTM neural network, comprising:

[0047] Data information acquisition module, for acquiring multiple die casting machine injection data information, the injection data information includes the injection speed of the die casting machine and the corresponding injection valve core displacement amount;

[0048] Inverse model establishment module, for combining the acquired multiple injection data information and the preset LSTM neural network model, and establishing the injection system LSTM neural network inverse model according to the preset model establishment algorithm;

[0049] Get inverse data module, for acquiring the preset injection valve core displacement amount corresponding to the preset injection speed of the die casting machine according to the injection system LSTM neural network inverse model;

[0050] Control die casting machine injection module, for controlling the corresponding die casting machine to inject according to the preset injection valve core displacement amount of the injection valve core movement.

[0051] Further, the inverse model establishment module further comprises a data preprocessing unit,

[0052] The data preprocessing unit is used for decomposing the injection speed data information according to the preset high-frequency speed information and the preset low-frequency speed information through wavelet decomposition; denoising the preset high-frequency speed information through wavelet threshold denoising algorithm; reconstructing the denoised data information according to the preset reconstruction algorithm; and acquiring the reconstructed injection speed curve information.

[0053] Further, it further comprises an iterative learning module,

[0054] The iterative learning module is used for acquiring the actual injection speed of the die casting machine when the corresponding die casting machine is controlled to inject according to the preset injection valve core displacement amount, and iteratively learning the actual injection speed according to the preset iterative learning algorithm when the actual injection speed is within the preset error range of the injection speed, until the actual injection speed of the die casting machine is within the preset error range of the injection speed.

[0055] Compared with the prior art, the present application has the following beneficial effects:

[0056] 1. By establishing the injection system LSTM neural network inverse model, the valve core displacement amount corresponding to the preset injection speed can be accurately predicted, and a more accurate initial injection speed and / or valve core displacement amount data input is provided for subsequent iterative learning, so that the iterative learning can achieve more effective control.

[0057] 2. The problem of large linear model error in the method of acquiring injection speed and valve core displacement amount by using the existing simple linear method is solved;

[0058] 3、The iteration learning in the application can effectively solve the problem that the experienced operator manually adjusts the displacement of the valve core and then makes the injection speed reach the expected value each time. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 A pressure injection principle diagram of a die casting machine;

[0060] Figure 2 An LSTM neural network model framework diagram of a pressure injection system;

[0061] Figure 3 A pressure injection curve diagram of a die casting machine;

[0062] Figure 4 A die casting machine pressure injection speed control method flowchart based on an LSTM neural network Figure 1 ;

[0063] Figure 5 A die casting machine pressure injection speed control method flowchart based on an LSTM neural network Figure 2 ;

[0064] Figure 6 A flowchart of iteration learning in an embodiment of a die casting machine pressure injection speed control method based on an LSTM neural network;

[0065] Figure 7 A die casting machine pressure injection speed control system based on an LSTM neural network Figure 1 .

[0066] 1, mold; 2, punch; 3, pressure injection cylinder; 4, booster cylinder; 5, magnetic grating displacement sensor; 6, first fast adjustment; 7, second fast adjustment; 8, accumulator. EMBODIMENTS

[0067] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0068] In the description of the present application, it should be understood that the terms "upper", "lower", etc. indicate the orientation or positional relationship based on the orientation or positional relationship when the product is normally used.

[0069] The terms "first", "second", etc. are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. Embodiment one

[0070] This invention provides a method for controlling the injection speed of a die-casting machine based on an LSTM neural network, such as... Figures 1 to 6 As shown, this method includes two parts: the establishment of an LSTM neural network inversion model for the injection system and the control of iterative learning of the injection speed. The establishment of the LSTM neural network inversion model for the injection system can accurately predict the valve core displacement corresponding to the preset injection speed. The iterative learning of the injection speed can further verify and iterate the preset data obtained by the LSTM neural network inversion model for the injection system, and obtain more accurate data information.

[0071] Specifically, the method for establishing the LSTM neural network inversion model of the pressure injection system includes the following steps:

[0072] A1: Obtain injection data information from multiple die-casting machines, including the injection speed of the die-casting machine and the corresponding displacement of the injection valve core.

[0073] During the injection process of the die-casting machine, the displacement-time information of the injection piston rod in the injection cylinder is collected by a preset displacement sensor, such as... Figure 1 As shown, the collected displacement-time information of the injection piston rod is further processed into corresponding injection velocity-time information.

[0074] The injection speed-time information from the die-casting machine forms an injection speed curve, which is a curve that changes with time, such as... Figure 3 As shown; in order to obtain the relationship between the injection speed of the die-casting machine and the displacement of the valve core, the present invention uses an LSTM neural network to perform data inversion; see step A2 for details.

[0075] A2: Combining the acquired multiple injection data and the preset LSTM neural network model, establish an LSTM neural network inversion model for the injection system according to the preset model establishment algorithm;

[0076] In this embodiment, the LSTM neural network inversion model of the injection system consists of two parts: data preprocessing and the LSTM neural network model. The LSTM neural network model includes a sequence input layer, an LSTM layer, a ReLU layer, a fully connected layer, and a regression layer, as shown in the figure. Figure 2 As shown.

[0077] The specific method for establishing the LSTM neural network inversion model of the pressure injection system is as follows:

[0078] Obtain the injection speed of multiple die-casting machines and the corresponding displacement of the injection valve core;

[0079] The injection speed of the die-casting machine is preprocessed to obtain the preprocessed injection speed data.

[0080] The injection speed data information after data preprocessing and the corresponding injection valve core displacement amount are input into a preset LSTM neural network model.

[0081] Further, the data preprocessing comprises the following steps:

[0082] The injection speed data information in the injection data information of the die casting machine is obtained.

[0083] The injection speed data information is decomposed into preset high-frequency speed information and preset low-frequency speed information through wavelet decomposition.

[0084] The preset high-frequency speed information is denoised through a wavelet threshold denoising algorithm.

[0085] The denoised data information is reconstructed according to a preset reconstruction algorithm.

[0086] The reconstructed injection speed data information is obtained.

[0087] The obtained injection speed data information may have different degrees of noise, and needs to be denoised before analyzing the injection speed data information (injection speed-time).

[0088] First, the injection speed information is decomposed into multiple high-frequency and multiple low-frequency components through wavelet decomposition, and then the corresponding high-frequency part is denoised through wavelet threshold denoising. The signal after wavelet threshold denoising is reconstructed, that is, the low-frequency wavelet coefficients after wavelet decomposition and the high-frequency wavelet coefficients updated by threshold denoising are inversely transformed by wavelet decomposition, so as to obtain an injection speed-time curve that can more accurately reflect the injection speed data information.

[0089] Iterative training is performed;

[0090] The injection speed data information under multiple different valve core displacements after the above data preprocessing is collected and input into the LSTM neural network model. Through repeated iterative training, the internal relationship between the injection speed of the injection system and the valve core displacement is saved in the weights and biases of the LSTM neural network, thereby completing the automatic learning of the internal relationship of the injection system.

[0091] An LSTM neural network inversion model of the injection system is established.

[0092] A3: Through the LSTM neural network inversion model of the injection system, the preset injection speed corresponding to the preset injection valve core displacement of the die casting machine is obtained.

[0093] A4: Control the injection valve core to move the preset injection valve core displacement and control the corresponding die casting machine to perform injection.

[0094] The injection speed data information of the expected injection of the die casting machine is input into the trained injection system LSTM neural network inversion model to obtain the corresponding valve core displacement amount, and finally the valve core displacement amount signal is input into the motor for driving the valve core to move, so that the motor drives the valve core to move the corresponding displacement, thereby realizing the injection at the corresponding speed.

[0095] The above steps are for establishing the injection system LSTM neural network inversion model. Through the establishment of the injection system LSTM neural network inversion model, the preset expected injection speed data information of the die casting machine can be input to obtain the corresponding valve core displacement amount, so as to control the motor for driving the valve core to move to the corresponding displacement amount, thereby further realizing the injection at the corresponding speed.

[0096] Since there is still a certain deviation between the actual injection speed and the expected speed output by the injection system LSTM neural network inversion model, an iterative learning algorithm is provided in the embodiment to correct the deviation. The specific iterative learning algorithm includes the following steps:

[0097] First, it is judged whether the deviation between the actual injection speed and the expected speed output by the injection system LSTM neural network inversion model is within the error range. If yes, the die casting machine can be continuously controlled according to the injection system LSTM neural network inversion model to obtain the corresponding valve core displacement amount according to the preset expected injection speed data information of the die casting machine; if not, the deviation is corrected by obtaining the corresponding valve core displacement amount according to the preset expected injection speed data information of the die casting machine.

[0098] A5: obtaining the actual injection speed of the die casting machine when the die casting machine is controlled according to the preset injection valve core displacement amount;

[0099] A6: judging whether the actual injection speed is within the preset error range;

[0100] The step A6 includes the following steps:

[0101] obtaining the preset actual speed difference between the preset injection speed and the actual injection speed;

[0102] judging whether the preset actual speed difference is within the preset error range;

[0103] If yes, the actual injection speed is within the preset error range;

[0104] If no, the actual injection speed is not within the preset error range.

[0105] A7: If yes, keep the actual injection speed corresponding to the preset injection valve core displacement control corresponding to the die casting machine for injection;

[0106] A8: If no, according to the preset iterative learning algorithm, the actual injection speed is iteratively learned until the actual injection speed of the die casting machine is within the preset error range.

[0107] The above steps are deviation correction for determining whether to perform iterative learning. If iterative learning is required, the following steps are performed:

[0108] The actual injection speed corresponding to the verification injection valve core displacement is obtained through the injection system LSTM neural network inversion model;

[0109] The displacement difference between the preset injection valve core displacement and the verification injection valve core displacement is obtained;

[0110] The obtained displacement difference is input into the preset proportional iterative learning algorithm, and the iterative injection valve core displacement is obtained;

[0111] The injection valve core is controlled to move the iterative injection valve core displacement, and the corresponding die casting machine is controlled for injection;

[0112] The actual injection speed of the die casting machine corresponding to the control of the injection valve core moving the iterative injection valve core displacement is obtained;

[0113] Step A6 is continued to be executed until the actual injection speed of the die casting machine is within the preset error range.

[0114] Specifically as shown in Figure 6 ,

[0115] A desired die casting machine preset injection speed is set, which is input into the injection system LSTM neural network inversion model to obtain the predicted valve core displacement X f1 ;

[0116] Then the valve core displacement X f1 signal is input to the motor controlling the valve core movement, and the motor drives the valve core to move to the required displacement X f1 , and then the electromagnetic valve on the plug-in valve controls the plug-in valve to open to realize the injection action;

[0117] And get the actual injection speed of the actual die casting machine, and then compare the actual injection speed with the also set injection to make a difference, and judge whether the speed is within the error range;

[0118] If the speed error is within the allowable range, the valve core displacement is kept unchanged for subsequent injection process;

[0119] If the speed error exceeds the allowable range, the actual injection speed size is input again into the injection system LSTM neural network inversion model to obtain a new valve core displacement, i.e. a verification injection valve core displacement, and the difference between the preset injection valve core displacement calculated in the foregoing and the verification injection valve core displacement is obtained as ΔX;

[0120] The difference ΔX is input into the proportional iterative learning algorithm to calculate a new valve core displacement X f2 , i.e. an iterative injection valve core displacement;

[0121] The iterative valve core displacement X f2 is input again to the motor, the motor drives the valve core to adjust, and then injection is performed again.

[0122] It is judged whether the injection speed after the re-injection is within the error allowable range. If the error is met, the iterative valve core displacement is kept unchanged for the next injection. If the error exceeds the error range, iterative operation is performed again to update the valve core displacement until the injection speed reaches the error allowable range.

[0123] The injection speed control method based on the LSTM neural network provided in the embodiment can accurately predict the valve core displacement corresponding to the expected injection speed, provide a more accurate initial input for subsequent iterative learning control, and enable the iterative learning control system to achieve more effective control. At the same time, the problem of large error of the linear model established in the prior art is solved, and the problem that an experienced operator manually adjusts the valve core displacement and then makes the injection speed reach the expected value each time is effectively solved. Embodiment Two

[0124] The embodiment provides an injection speed control system for a die casting machine based on an LSTM neural network. The system includes the control method in Embodiment One, as shown in Figure 7 The system includes a data information acquisition module for acquiring injection data information of multiple die casting machines, the injection data information including injection speed and corresponding injection valve core displacement of the die casting machine;

[0125] An inversion model establishment module is configured to combine the acquired multiple injection data information and a preset LSTM neural network model, and establish a die casting system LSTM neural network inversion model according to a preset model establishment algorithm;

[0126] An inversion data acquisition module is configured to acquire a preset injection valve core displacement corresponding to a preset injection speed of the die casting machine according to the die casting system LSTM neural network inversion model;

[0127] A control die casting machine injection module is configured to control the corresponding die casting machine to perform injection according to the preset injection valve core displacement of the injection valve core. The fast stage of the injection of the die casting machine is divided into a first fast state and a second fast state, as shown in the figure, which are controlled by two plug valves respectively. The first fast speed is the speed provided by the flow when the first plug valve is opened, and the second fast speed is the speed provided by the flow when the first plug valve and the second plug valve are opened simultaneously. The speed during the second fast injection can be considered as the superposition of the flow provided by the opening of the first plug valve and the second plug valve.

[0128] Figure 1 Therefore, the injection system LSTM neural network inversion model of the valve core displacement amount corresponding to the first fast plug valve and the second fast plug valve when they are opened alone can be established respectively, and then the prediction of the valve core displacement amount corresponding to the first fast speed is predicted by the model corresponding to the first fast plug valve; the prediction of the second fast speed can be predicted by the valve core displacement amount corresponding to the second fast plug valve when the second fast speed is subtracted from the first fast speed as the input of the model corresponding to the second fast plug valve.

[0129] Therefore, the injection system LSTM neural network inversion model of the valve core displacement amount corresponding to the first fast plug valve and the second fast plug valve when they are opened alone can be established respectively, and then the prediction of the valve core displacement amount corresponding to the first fast speed is predicted by the model corresponding to the first fast plug valve; the prediction of the second fast speed can be predicted by the valve core displacement amount corresponding to the second fast plug valve when the second fast speed is subtracted from the first fast speed as the input of the model corresponding to the second fast plug valve.

[0130] The inversion model establishing module further comprises a data preprocessing unit, which is used for decomposing the injection speed data information according to preset high-frequency speed information and preset low-frequency speed information through wavelet decomposition; denoising the preset high-frequency speed information through a wavelet threshold denoising algorithm; reconstructing the denoised data information according to a preset reconstruction algorithm; and obtaining the reconstructed injection speed data information.

[0131] The inversion model establishing module further comprises an iterative learning module, which is used for obtaining the actual injection speed of the die casting machine when the die casting machine is controlled according to the preset injection valve core displacement amount, and iteratively learning the actual injection speed according to a preset iterative learning algorithm when the actual injection speed is within the preset error range, until the actual injection speed of the die casting machine is within the preset error range.

[0132] The system can automatically learn the internal relationship between the injection speed and the valve core displacement in the injection system through the establishment of the injection system LSTM neural network inversion model, and save the internal relationship in the weights and biases of the LSTM neural network. The established injection system LSTM neural network inversion model can accurately describe the relationship between the injection speed and the valve core displacement, so as to effectively solve the problem of large error in predicting the valve core displacement by the existing linear model.

[0133] On the basis of the established injection system LSTM neural network inversion model, a desired speed (preset injection speed) corresponding valve core displacement can be initially predicted, the signal of the displacement is input to the motor controlling the valve core movement, the motor drives the valve core to move to the specified value, and then injection is carried out. The value of the injection speed is fed back to the iterative learning algorithm for updating the valve core displacement. The updated valve core displacement signal is input to the motor again, so as to adjust the valve core displacement and perform re-injection, thereby completing an iterative control until the injection speed reaches the error range to end the iterative operation. The designed iterative learning speed control method can automatically reach the expected value after completing the preset number of injections, and can effectively solve the problem that the valve core displacement needs to be adjusted manually by experienced operators every time, and then the injection speed reaches the preset injection speed.

[0134] Although the present disclosure is disclosed as above, the protection scope of the present disclosure is not limited to this. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present disclosure, and these changes and modifications shall fall within the protection scope of the present disclosure.

Claims

1. A method for controlling the injection speed of a die-casting machine based on an LSTM neural network, characterized in that, Including the following steps: A1: Obtain injection data information from multiple die-casting machines, including the injection speed of the die-casting machine and the corresponding displacement of the injection valve core; A2: Combining the acquired multiple injection data and the preset LSTM neural network model, establish an LSTM neural network inversion model for the injection system according to the preset model establishment algorithm; A3: Obtain the preset injection valve core displacement corresponding to the preset injection speed of the die-casting machine by using the LSTM neural network inversion model of the injection system; A4: Controls the movement of the injection valve core by a preset displacement and controls the corresponding die-casting machine to perform injection; whereby... The steps of establishing an LSTM neural network inversion model of the injection system by combining multiple injection data and a preset LSTM neural network model according to a preset model establishment algorithm include: Obtain the injection speed of multiple die-casting machines and the corresponding displacement of the injection valve core; The injection speed of the die-casting machine is preprocessed to obtain the preprocessed injection speed data. The preprocessed injection velocity data and the corresponding injection valve core displacement are input into a preset LSTM neural network model; Perform iterative training; Establish an LSTM neural network inversion model for the pressure injection system; The data preprocessing includes the following steps: Obtain the injection speed data from the injection data information of the die-casting machine; The injection velocity data is decomposed into preset high-frequency velocity information and preset low-frequency velocity information by wavelet decomposition. The preset high-frequency speed information is denoised using a wavelet threshold denoising algorithm; The denoised data is reconstructed according to the preset reconstruction algorithm; Obtain the reconstructed injection velocity data.

2. The injection speed control method for a die-casting machine based on an LSTM neural network according to claim 1, characterized in that, It also includes the following steps: A5: Obtain the actual injection speed of the die-casting machine when it performs injection according to the preset injection valve core displacement; A6: Determine whether the actual injection speed is within the preset error range; A7: If so, continue to maintain the preset displacement of the injection valve core corresponding to the actual injection speed to control the corresponding die-casting machine to perform injection; A8: If not, the actual injection speed is iteratively learned according to the preset iterative learning algorithm until the actual injection speed of the die casting machine is within the preset error range.

3. The injection speed control method for a die-casting machine based on an LSTM neural network according to claim 2, characterized in that, Determining whether the actual injection velocity is within the preset error range includes the following steps: Obtain the preset injection speed and the preset-actual speed difference between the preset injection speed and the actual injection speed; Determine whether the preset actual speed difference is within the preset error range; If so, the actual injection speed is the injection speed within the preset error range; If not, the actual injection speed is not within the preset error range.

4. The injection speed control method for a die-casting machine based on an LSTM neural network according to claim 2, characterized in that, The step of iteratively learning the actual injection speed according to a preset iterative learning algorithm until the actual injection speed of the die-casting machine is within a preset error range includes the following steps: The displacement of the verification injection valve core corresponding to the actual injection speed is obtained by using the LSTM neural network inversion model of the injection system. Obtain the displacement difference between the preset injection valve core displacement and the verified injection valve core displacement; The obtained displacement difference is input into a preset proportional iterative learning algorithm to obtain the iterative injection valve core displacement. Control the movement of the injection valve core, iterate the displacement of the injection valve core, and control the corresponding die-casting machine to perform injection; Obtain the actual injection speed of the die-casting machine corresponding to the iterative displacement of the control injection valve core; Continue with step A6 until the actual injection speed of the die-casting machine is within the preset error range.

5. The injection speed control method for a die-casting machine based on an LSTM neural network according to claim 4, characterized in that, The preset proportional iterative learning algorithm is as follows: X f2 =X f1 +K△X Where X f2 For: Iterative injection valve core displacement; Where X f1 For: preset injection valve core displacement; Where △X is the displacement difference between the preset injection valve core displacement and the verified injection valve core displacement; K is the proportionality coefficient.

6. A die-casting machine injection speed control system based on LSTM neural network, characterized in that, include: The data information acquisition module is used to acquire injection data information from multiple die-casting machines. The injection data information includes the injection speed of the die-casting machine and the corresponding displacement of the injection valve core. The inversion model building module is used to combine the acquired multiple injection data and the preset LSTM neural network model, and build the LSTM neural network inversion model of the injection system according to the preset model building algorithm. The inversion data acquisition module is used to obtain the preset injection valve core displacement corresponding to the preset injection speed of the die casting machine based on the LSTM neural network inversion model of the injection system. The injection control module of the die casting machine is used to control the corresponding die casting machine to perform injection based on the preset displacement of the injection valve core. The inversion model building module also includes a data preprocessing unit. The data preprocessing unit is used to decompose the injection velocity data information into preset high-frequency velocity information and preset low-frequency velocity information through wavelet decomposition; to denoise the preset high-frequency velocity information through a wavelet threshold denoising algorithm; to reconstruct the denoised data information according to a preset reconstruction algorithm; and to obtain the reconstructed injection velocity curve information.

7. The die-casting machine injection speed control system based on LSTM neural network according to claim 6, characterized in that, It also includes an iterative learning module. The iterative learning module is used to obtain the actual injection speed of the die-casting machine when it is controlled by the preset injection valve core displacement. When the actual injection speed is not within the preset error range, the module iteratively learns the actual injection speed according to the preset iterative learning algorithm until the actual injection speed of the die-casting machine is within the preset error range.

Images