Dispensing valve input voltage correction method, device, equipment and medium

Abstract

The application provides a dispensing valve input voltage correction method, device, equipment and medium, and belongs to the technical field of dispensing valves. The method comprises the following steps: drawing a corresponding relationship diagram, obtaining a plurality of training samples by using the corresponding relationship diagram, training a neural network model by using the training samples, and obtaining a prediction model. An input voltage is input into the prediction model to obtain an actual output displacement corresponding voltage, and a hysteresis displacement is calculated. By using the specific curve law between the input voltage and the output displacement in the lift and return processes, the to-be-tested input voltage is input into the trained prediction model to obtain an actual output displacement corresponding voltage eliminating the hysteresis effect, and the actual output displacement corresponding voltage is used as the actual input voltage of the dispensing valve. The above-mentioned method can eliminate the influence of hysteresis, and the to-be-tested input voltage and the actual output displacement corresponding voltage affected by hysteresis are corresponded to each other by using the prediction model, so that the input voltage is corrected, and the output displacement of the dispensing valve is more accurately controlled.

Description

Technical Field

[0001] This invention generally relates to the field of dispensing valve technology, and specifically to a method, apparatus, equipment, and medium for correcting the input voltage of a dispensing valve. Background Technology

[0002] Dispensing technology is the process of distributing fluids to designated locations to achieve functions such as fixing, encapsulating, and soldering electronic components. When electronic components need to be connected to electrical appliances, dispensing technology can distribute solder paste, silver paste, etc., to designated locations to achieve electrical connection between electronic components and electrical components. Electronic components face various complex situations in actual operation, and dispensing technology can also encapsulate chips, giving them functions such as waterproofing, electrical protection, moisture protection, impact protection, and corrosion protection.

[0003] The impact-type spray dispensing valve is normally closed when the machine is stopped, with the impact pin and nozzle in a tight seal. When the driving force lifts the impact pin (lifting process), the pin disengages from the nozzle, and the adhesive in the flow channel is transported to the gap between the impact pin and the nozzle under pressure. When the impact pin strikes the nozzle downwards (return process), a momentary high pressure is created between the impact pin and the nozzle. This high pressure forces the adhesive to flow rapidly out of the nozzle orifice; simultaneously, when the impact pin strikes the nozzle, it cuts off the flow of adhesive from the flow channel to the nozzle. This momentary high pressure forces the adhesive in the nozzle orifice to break free from the fluid force at the nozzle, thus achieving spraying.

[0004] Specifically, both the lift and return processes of the dispensing valve are controlled by the input voltage to generate output displacement of the valve body. Ideally, without considering the effect of hysteresis, the required output displacement can be calculated by substituting the input voltage into the linear displacement model of the dispensing valve.

[0005] However, in actual dispensing valve control, a hysteresis effect is usually present, causing the corresponding input voltage to fail to produce the required output displacement. During the lift phase, the actual output displacement is always less than the required output displacement; during the return phase, the actual output displacement is always greater than the required output displacement. Therefore, the actual output displacement of the dispensing valve has a significant error compared to the calculated required displacement, resulting in uneven dispensing of the adhesive and a large difference between the dispensing volume and the actual required volume. Therefore, it is necessary to correct the input voltage to ensure the required output displacement. Summary of the Invention

[0006] In view of the above-mentioned defects or deficiencies in the prior art, it is desirable to provide a method, apparatus, device and medium for correcting the input voltage of a dispensing valve.

[0007] In a first aspect, the present invention provides a method for correcting the input voltage of a dispensing valve considering the effect of hysteresis, comprising:

[0008] Obtain the correspondence diagram; the correspondence diagram includes multiple sample data points consisting of the input voltage and output displacement of the dispensing valve during the lift and return processes, as well as the Bode plot corresponding to the linear model of the dispensing valve displacement;

[0009] Multiple training samples are obtained based on the correspondence diagram; the training samples include input voltage and output displacement corresponding voltage; the output displacement corresponding voltage is calculated by substituting the output displacement corresponding to the input voltage into the linear displacement model of the dispensing valve.

[0010] Obtain the neural network model;

[0011] The training samples are input into the neural network model, with the input voltage as the input and the voltage corresponding to the output displacement as the output, to train the neural network model and obtain a prediction model.

[0012] Obtain the input voltage to be predicted;

[0013] Input the input voltage to be predicted into the prediction model, predict the voltage corresponding to the output displacement, and obtain the voltage corresponding to the actual output displacement;

[0014] The voltage corresponding to the actual output displacement is used as the actual input voltage of the dispensing valve.

[0015] According to the technical solution provided by the present invention, the steps for obtaining the correspondence graph include:

[0016] Obtain the linear model of the dispensing valve displacement;

[0017] The sample data is obtained, including elevation sample data and return sample data;

[0018] The transfer function model and the sample data are plotted on the same coordinate system, with the input voltage as the horizontal axis and the output displacement as the vertical axis, to obtain the corresponding relationship diagram.

[0019] According to the technical solution provided by the present invention, the dispensing valve includes: a piezoelectric stack, a lever, and a striking pin assembly;

[0020] The method for constructing the linear model of the dispensing valve displacement includes:

[0021] The dynamic differential equations of the piezoelectric stack, the lever, and the firing pin assembly are established respectively, resulting in the differential equations of the piezoelectric stack, the lever, and the firing pin assembly.

[0022] Establish conversion equations between the piezoelectric stack's motion displacement, the lever's rotation angle, and the striker assembly's motion displacement;

[0023] Based on the differential equations of the piezoelectric stack, the lever, and the firing pin assembly, as well as the conversion equation, the dynamic differential equation of the dispensing valve is calculated.

[0024] The linear displacement model of the dispensing valve is calculated based on the dynamic differential equation of the dispensing valve.

[0025] According to the technical solution provided by the present invention, the step of obtaining multiple training samples based on the correspondence graph includes:

[0026] S41: Randomly select a point containing an unselected sample data according to the correspondence diagram to obtain the first sample point;

[0027] S42: Obtain the x-coordinate and y-coordinate of the first sample point to get the first sample input voltage and the first sample output displacement;

[0028] S43: Substitute the output displacement of the first sample into the linear displacement model of the dispensing valve to calculate the voltage corresponding to the output displacement;

[0029] S44: Determine the magnitude of the voltage corresponding to the output displacement and the first sample input voltage. When the voltage corresponding to the output displacement is less than the first sample input voltage, take the voltage corresponding to the output displacement and the first sample input voltage as a set of rise samples; when the voltage corresponding to the output displacement is greater than the first sample input voltage, take the voltage corresponding to the output displacement and the first sample input voltage as a set of return samples.

[0030] S45: Repeat steps S41-S44 until the number of lift samples and the number of return samples both reach the set total number of samples; thus obtaining multiple lift samples and multiple return samples.

[0031] S46: Use the multiple elevation samples and the multiple return samples as multiple training samples.

[0032] According to the technical solution provided by the present invention, the training samples are input into the neural network model, and before training the neural network model, the method further includes screening the training samples, specifically including:

[0033] The values ​​of the voltages corresponding to the multiple lift displacements for each lift input voltage are statistically analyzed, and the distribution curves of the voltages corresponding to the multiple lift displacements for each lift input voltage are plotted to obtain multiple distribution curves.

[0034] Calculate the expected value and standard deviation of each of the distribution curves;

[0035] The screening rules are set based on the aforementioned mathematical expectation and standard deviation;

[0036] The sample data is filtered according to the filtering rules.

[0037] The remaining sample data after filtering will be used as training samples.

[0038] According to the technical solution provided by the present invention, the neural network model includes: a first neural network model and a second neural network model;

[0039] The steps for training the neural network model to obtain the prediction model include:

[0040] The lift sample is input into the first neural network model, with the lift input voltage as input and the voltage corresponding to the lift displacement as output, and the first neural network model is trained to obtain the lift prediction model;

[0041] The return sample is input into the second neural network model, with the return input voltage as input and the voltage corresponding to the return displacement as output, to train the second neural network model and obtain the return prediction model;

[0042] The lift prediction model and the return prediction model are used together as the prediction model.

[0043] According to the technical solution provided by the present invention, the specific steps for inputting the input voltage to be predicted into the prediction model, predicting the voltage corresponding to the output displacement, and obtaining the voltage corresponding to the actual output displacement include:

[0044] Determine the movement state of the dispensing valve.

[0045] When the dispensing valve is in the lift state, the input voltage to be predicted is input into the lift prediction model to obtain the lift prediction voltage value; the lift prediction voltage value is used as the voltage corresponding to the actual output displacement.

[0046] When the dispensing valve is in the return state, the input voltage to be predicted is input into the return prediction model to obtain the return prediction voltage value; the return prediction voltage value is used as the voltage corresponding to the actual output displacement.

[0047] In a second aspect, the present invention provides a dispensing valve input voltage correction device based on a neural network model, used to perform a dispensing valve input voltage correction method considering hysteresis effects as described above, comprising:

[0048] Data storage module, the data storage module being used for:

[0049] The system stores multiple sample data consisting of the input voltage and output displacement of the dispensing valve during the lift and return processes, a linear model of the dispensing valve displacement, and a neural network model.

[0050] The processing module is used for:

[0051] Obtain the corresponding relationship graph;

[0052] Multiple training samples are obtained based on the correspondence graph;

[0053] The training samples are input into the neural network model, with the input voltage as the input and the voltage corresponding to the output displacement as the output, to train the neural network model and obtain a prediction model.

[0054] Obtain the input voltage to be predicted;

[0055] Input the input voltage to be predicted into the prediction model, predict the voltage corresponding to the output displacement, and obtain the voltage corresponding to the actual output displacement;

[0056] The hysteresis displacement is calculated based on the voltage corresponding to the actual output displacement and the linear model of the dispensing valve displacement.

[0057] In a third aspect, the present invention provides a computer device, comprising: a memory, a processor, and a program stored in the memory for performing input voltage correction of a dispensing valve; the program for performing input voltage correction of the dispensing valve is configured to:

[0058] The above describes a method for correcting the input voltage of a dispensing valve that takes into account the effects of hysteresis.

[0059] Fourthly, the present invention provides a storage medium storing a program for measuring the hysteresis displacement of a dispensing valve; when the program for measuring the hysteresis displacement of the dispensing valve is executed, it is used to:

[0060] The above describes a method for correcting the input voltage of a dispensing valve that takes into account the effects of hysteresis.

[0061] The beneficial effects of this invention are as follows:

[0062] By repeatedly measuring the input voltage and output displacement during the lift and return processes, a correspondence diagram is plotted using a linear displacement model of the dispensing valve. Multiple training samples are then obtained from this diagram, and the neural network model is trained using these samples to obtain a prediction model. The input voltage is fed into the prediction model to obtain the voltage corresponding to the actual output displacement, and the hysteresis displacement is calculated. Utilizing the specific curve relationship between the input voltage and output displacement during the lift and return processes, the measured input voltage is input into the trained prediction model to obtain the voltage corresponding to the actual output displacement after eliminating the hysteresis effect. This voltage corresponding to the actual output displacement is then used as the actual input voltage of the dispensing valve. This method eliminates the influence of hysteresis, and by using the prediction model to correlate the measured input voltage with the voltage corresponding to the actual output displacement after the hysteresis effect, the input voltage can be corrected, resulting in more precise control of the dispensing valve's output displacement. Attached Figure Description

[0063] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0064] Figure 1 A flowchart illustrating a method for correcting the input voltage of a dispensing valve considering the effects of hysteresis, provided by the present invention.

[0065] Figure 2 This is a simplified structural diagram of a dispensing valve;

[0066] Figure 3 A simplified geometric model diagram of a dispensing valve;

[0067] Figure 4 This is a schematic diagram of the correspondence relationship;

[0068] Figure 5 A schematic diagram for calculating training samples;

[0069] Among them: 1. Piezoelectric stack; 2. Lever; 3. Strike pin assembly. Detailed Implementation

[0070] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0071] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0072] Example 1

[0073] Please refer to Figure 1 The flowchart illustrates a method for correcting the input voltage of a dispensing valve considering hysteresis, as provided by the present invention, including:

[0074] Obtain the correspondence diagram; the correspondence diagram includes multiple sample data points consisting of the input voltage and output displacement of the dispensing valve during the lift and return processes, as well as the Bode plot corresponding to the linear model of the dispensing valve displacement;

[0075] Multiple training samples are obtained based on the correspondence diagram; the training samples include input voltage and output displacement corresponding voltage; the output displacement corresponding voltage is calculated by substituting the output displacement corresponding to the input voltage into the linear displacement model of the dispensing valve.

[0076] Obtain the neural network model;

[0077] The training samples are input into the neural network model, with the input voltage as the input and the voltage corresponding to the output displacement as the output, to train the neural network model and obtain a prediction model.

[0078] Obtain the input voltage to be predicted;

[0079] Input the input voltage to be predicted into the prediction model, predict the voltage corresponding to the output displacement, and obtain the voltage corresponding to the actual output displacement;

[0080] The voltage corresponding to the actual output displacement is used as the actual input voltage of the dispensing valve.

[0081] In some implementations, such as Figure 4 As shown, the correspondence diagram includes: multiple sample data consisting of the input voltage and output displacement of the dispensing valve during the lifting process, and multiple sample data consisting of the input voltage and output displacement of the dispensing valve during the return process.

[0082] Specifically, Figure 4 In this diagram, the x-axis represents the input voltage and the y-axis represents the output displacement for each data point. Figure 4 The sloping straight line represents the Bode plot corresponding to the linear model G(S) of the dispensing valve displacement. Under the same horizontal axis, the sample data of the data points with a vertical axis greater than that of the linear model of the dispensing valve displacement belong to the sample data of the return process, and the sample data of the data points with a vertical axis less than that of the linear model of the dispensing valve displacement belong to the sample data of the lift process.

[0083] The sample data is input into the neural network model, and the neural network model is used to find the correspondence between the input voltage and the output displacement under the hysteresis effect.

[0084] by Figure 4Taking the data points in the model as an example, during the lift process, if the input voltage to be predicted is E0, and the actual input displacement is less than the first output displacement calculated according to the linear model of the dispensing valve displacement, then the coordinates of the sample data whose output displacement is equal to the first output displacement are found, and the first input voltage E1 corresponding to that sample data is obtained. If the input voltage to be predicted is E0 during the lift process, then the first input voltage E1 is taken as the actual input value.

[0085] During the return stroke, if the input voltage to be predicted is D0, and the actual input displacement is greater than the second output displacement calculated according to the linear model of the dispensing valve displacement, then the coordinates of the sample data where the output displacement is equal to the second output displacement are found, and the second input voltage D1 corresponding to that sample data is obtained. If the input voltage to be predicted is D0 during the return stroke, then the second input voltage D1 is taken as the actual input value.

[0086] In some implementations, a large amount of sample data is collected in accordance with the above correspondence, and the neural network model is trained using the input voltage to be predicted and the voltage corresponding to the actual output displacement, so that the neural network model becomes a prediction model for predicting the voltage corresponding to the actual output displacement based on the input voltage to be predicted.

[0087] In the actual prediction process, the input voltage to be predicted is input into the prediction model, and the voltage corresponding to the actual output displacement is obtained through prediction. Therefore, the voltage corresponding to the actual output displacement can be directly used as the input voltage to adjust to the required output displacement.

[0088] The above method can eliminate the effects of hysteresis. By using a predictive model to match the input voltage to be measured with the voltage corresponding to the actual output displacement after the hysteresis effect, the input voltage can be corrected, and the output displacement of the dispensing valve can be controlled more accurately.

[0089] Further, the step of obtaining the correspondence graph includes:

[0090] Obtain the linear model of the dispensing valve displacement;

[0091] The sample data is obtained, including elevation sample data and return sample data;

[0092] The transfer function model and the sample data are plotted on the same coordinate system, with the input voltage as the horizontal axis and the output displacement as the vertical axis, to obtain the corresponding relationship diagram.

[0093] Specifically, the lift sample data is in Figure 4 The value is represented by E, which contains sample data from multiple lift processes. Figure 4 The data is displayed as multiple data points; the return sample data is in Figure 4 The value in the middle is represented by D, which contains sample data from multiple return processes. Figure 4 The data is displayed as multiple data points.

[0094] In some implementations, the data points are obtained through experimental measurements. The data acquisition process includes: during the lift-off phase, multiple output displacements are acquired for the same input voltage, and the same acquisition process is repeated for different input voltages. During the return phase, multiple output displacements are acquired for the same input voltage, and the same acquisition process is repeated for different input voltages.

[0095] Due to various errors, the output displacements measured multiple times under the same input voltage are not exactly the same. Statistically, multiple output displacements collected under the same input voltage follow a normal distribution.

[0096] The collected data points and the linear model of the dispensing valve displacement were plotted in the same coordinate system and on the same horizontal axis. The data points during the lift process were all smaller than the data points of the linear model of the dispensing valve displacement, while the data points during the return process were all larger than the data points of the linear model of the dispensing valve displacement.

[0097] Furthermore, the dispensing valve includes: a piezoelectric stack 1, a lever 2, and a firing pin assembly 3;

[0098] The method for constructing the linear model of the dispensing valve displacement includes:

[0099] The dynamic differential equations of the piezoelectric stack 1, the lever 2, and the firing pin assembly 3 are established respectively, resulting in the differential equations of the piezoelectric stack, the lever, and the firing pin assembly.

[0100] Establish conversion equations between the motion displacement of piezoelectric stack 1, the rotation angle of lever 2, and the motion displacement of striker assembly 3;

[0101] Based on the differential equations of the piezoelectric stack, the lever, and the firing pin assembly, as well as the conversion equation, the dynamic differential equation of the dispensing valve is calculated.

[0102] The linear displacement model of the dispensing valve is calculated based on the dynamic differential equation of the dispensing valve.

[0103] Specifically, refer to Figure 2 and Figure 3 , Figure 2 In this paper, the piezoelectric stack 1 is simplified into a simple model of a spring and a piezoelectric stack counterweight, and the striker assembly 3 is simplified into a simple model of a striker, a striker counterweight, a large spring and a small spring.

[0104] Figure 3 middle, F 1 represents the force exerted by piezoelectric stack 1 on lever 2. a This represents the distance between the rotation center of lever 2 and the point of action of piezoelectric stack 1 on lever 2. bThis indicates the distance between the rotation center of lever 2 and the firing pin assembly 3. x 1 represents the elongation displacement of the piezoelectric stack 1. x 2 indicates that the firing pin 2 has moved and shifted. θ This indicates the rotation angle of lever 2.

[0105] In some implementations, the calculation process for constructing a linear model of the dispensing valve displacement includes:

[0106] Establish the differential equation of lever 2 according to formula (1);

[0107] Formula (1);

[0108] in, I This represents the moment of inertia of lever 2 about its center of rotation. This represents the angular acceleration of lever 2. F 1 represents the force exerted by piezoelectric stack 1 on lever 2. F 2 represents the force exerted by lever 2 on the firing pin. a This represents the distance between the rotation center of lever 2 and the point of action of piezoelectric stack 1 on lever 2. b This indicates the distance between the rotation center of lever 2 and the firing pin assembly 3.

[0109] The differential equation of the piezoelectric stack 1 is established according to formula (II);

[0110] Formula (II);

[0111] in, This represents the equivalent mass of the piezoelectric stack 1. This represents the acceleration of the piezoelectric stack 1. f This indicates the output force of the piezoelectric stack 1. F 1 represents the force exerted by piezoelectric stack 1 on lever 2. x 1 represents the elongation displacement of the piezoelectric stack 1. This is expressed as the stiffness of a piezoelectric stack. d 33 This represents the piezoelectric constant of the piezoelectric element. n This represents the total number of piezoelectric elements that make up piezoelectric stack 1. u This indicates the voltage applied across the piezoelectric element.

[0112] The striking pin assembly 3 includes: a striking pin, a small spring, and a large spring. The differential equation of the striking pin assembly 3 is established according to formula (III).

[0113] Formula (3);

[0114] in, Indicates the mass of the firing pin. Indicates the stiffness of the small spring. Indicates the stiffness of a large spring. x 2 indicates that the firing pin has moved and shifted. Indicates the speed of the firing pin. This represents the acceleration of the firing pin. F 2 represents the force exerted by lever 2 on the striking pin.

[0115] In the actual dispensing valve transmission process, the rotation angle of lever 2 is very small, and the following approximate relationship can be established, as shown in formula (IV);

[0116] Formula (IV);

[0117] in, a This represents the distance between the rotation center of lever 2 and the point of action of piezoelectric stack 1 on lever 2. b This indicates the distance between the rotation center of lever 2 and the firing pin assembly 3. x 1 represents the elongation displacement of the piezoelectric stack 1. x 2 indicates that the firing pin has moved and shifted. θ This indicates the rotation angle of lever 2.

[0118] Furthermore, based on formula (iv), we can deduce that:

[0119] Formula (5);

[0120] Formula (VI);

[0121] in, x 1 represents the elongation displacement of the piezoelectric stack 1. This indicates the velocity of piezoelectric stack 1. This represents the acceleration of piezoelectric stack 1; x 2 indicates that the firing pin has moved and shifted. Indicates the speed of the firing pin. This indicates the acceleration of the firing pin; a This represents the distance between the rotation center of lever 2 and the point of action of piezoelectric stack 1 on lever 2. b This indicates the distance between the rotation center of lever 2 and the firing pin assembly 3; θ This indicates the rotation angle of lever 2. This represents the rotational angular velocity of lever 2. This represents the rotational angular acceleration of lever 2.

[0122] Formulas (iv), (v) and (vi) are used as conversion equations between the displacement of the piezoelectric stack 1, the rotation angle of the lever 2, and the displacement of the striker assembly 3.

[0123] By combining equations (I), (II), (III), (IV), (V), and (VI), the differential equation of motion for the dispensing valve is obtained, as shown in equation (VII):

[0124]

[0125] Formula (VII);

[0126] in, a This represents the distance between the rotation center of lever 2 and the point of action of piezoelectric stack 1 on lever 2. b This indicates the distance between the rotation center of lever 2 and the firing pin assembly 3. Indicates the mass of the firing pin. This represents the equivalent mass of the piezoelectric stack 1. c Indicates system damping. Indicates the stiffness of the small spring. Indicates the stiffness of a large spring. This is expressed as the stiffness of a piezoelectric stack. x 2 indicates that the firing pin has moved and shifted. Indicates the speed of the firing pin. This represents the acceleration of the firing pin. d 33 This represents the piezoelectric constant of the piezoelectric element. n This represents the total number of piezoelectric elements that make up piezoelectric stack 1. u This indicates the voltage applied across the piezoelectric element. I This represents the moment of inertia of lever 2 about its center of rotation.

[0127] By performing a Laplace transform on both sides of Equation (VII), the transfer function of the dispensing valve system is obtained, as shown in Equation (VIII).

[0128]

[0129] Formula (8);

[0130] in, G(S) The system function of the dispensing valve, i.e., the linear model of the dispensing valve displacement, represents... X 2 (S) Indicates the system output. U(S) The input to the system is represented by s, which is a complex variable representing the frequency in the complex plane.

[0131] make , , ,

[0132] The dispensing valve system function can be obtained, as shown in formula (IX);

[0133] Formula (IX);

[0134] in, G(S) This represents the function of the dispensing valve system. X 2 (S) Indicates the system output. U(S) Indicates the system input, a 0、 a 1. a 2. b 0 represents system parameters.

[0135] The dispensing valve system is converted into a standard form to obtain a linear model of the dispensing valve displacement, as shown in formula (x).

[0136] Formula (10);

[0137] in, , Indicates the system damping ratio. This represents the undamped natural frequency.

[0138] Furthermore, the step of obtaining multiple training samples based on the correspondence graph includes:

[0139] S41: Randomly select a point containing an unselected sample data according to the correspondence diagram to obtain the first sample point;

[0140] S42: Obtain the x-coordinate and y-coordinate of the first sample point to get the first sample input voltage and the first sample output displacement;

[0141] S43: Substitute the output displacement of the first sample into the linear displacement model of the dispensing valve to calculate the voltage corresponding to the output displacement;

[0142] S44: Determine the magnitude of the voltage corresponding to the output displacement and the first sample input voltage. When the voltage corresponding to the output displacement is less than the first sample input voltage, take the voltage corresponding to the output displacement and the first sample input voltage as a set of rise samples; when the voltage corresponding to the output displacement is greater than the first sample input voltage, take the voltage corresponding to the output displacement and the first sample input voltage as a set of return samples.

[0143] S45: Repeat steps S41-S44 until the number of lift samples and the number of return samples both reach the set total number of samples; thus obtaining multiple lift samples and multiple return samples.

[0144] S46: Use the multiple elevation samples and the multiple return samples as multiple training samples.

[0145] In some implementations, the process of collecting training samples involves substituting the actual output displacement measured under the measured input voltage into the linear displacement model of the dispensing valve, and then calculating the standard input voltage under the condition of no hysteresis. After repeating this process multiple times, a correspondence between the standard input voltage and the actual input voltage can be established while ensuring that the output displacement is equal.

[0146] In the actual prediction process, the input voltage calculated by other methods without considering the hysteresis effect is used as the standard voltage. The prediction model can then predict the actual input voltage required to achieve the desired output displacement.

[0147] In some implementations, reference Figure 4 The training samples include lift samples E and return samples D; both lift and return samples contain multiple different input voltages.

[0148] In the lift sample E, each input voltage corresponds to multiple output displacements. Each output displacement is substituted into the linear displacement model of the dispensing valve to calculate the voltage corresponding to each lift displacement. Each voltage corresponding to a lift displacement is combined with the input voltage as training data to obtain multiple lift samples.

[0149] In the return sample D, each input voltage corresponds to multiple output displacements. Each output displacement is substituted into the linear displacement model of the dispensing valve to calculate the voltage corresponding to each return displacement. Each voltage corresponding to a return displacement is combined with the input voltage as training data to obtain multiple return samples.

[0150] Specifically, refer to Figure 5 Point A is randomly selected as the first sample point, with its abscissa U1 and ordinate X. X is substituted into the linear displacement model G(S) of the dispensing valve, i.e., formula (x), to calculate the voltage U0 corresponding to the output displacement, which is the abscissa of point B. Point B is located on the Bode plot of the linear displacement model of the dispensing valve. In this embodiment, since the voltage U0 corresponding to the output displacement is less than the abscissa of the first sample point U1, U0 and U1 are used as a set of lift samples.

[0151] In some implementations, the total number of samples is set as follows: within the input range of the input voltage, the total number of different input voltage values ​​is not less than 100; and the output displacement corresponding to the same input voltage is not less than 10.

[0152] Furthermore, before training the neural network model, the training samples are input into the neural network model, and the training samples are screened. Specific steps include:

[0153] The values ​​of the voltages corresponding to the multiple lift displacements for each lift input voltage are statistically analyzed, and the distribution curves of the voltages corresponding to the multiple lift displacements for each lift input voltage are plotted to obtain multiple distribution curves.

[0154] Calculate the expected value μ and standard deviation σ for each of the distribution curves;

[0155] The screening rules are set based on the aforementioned mathematical expectation and standard deviation;

[0156] The sample data is filtered according to the filtering rules.

[0157] The remaining sample data after filtering will be used as training samples.

[0158] In some implementations, since the same input voltage corresponds to multiple output displacements, and different models of dispensing valves require different adjustment accuracies, the training samples need to be screened according to certain screening rules before training the neural network model.

[0159] According to the statistical results, the multiple output displacements corresponding to the same input voltage basically follow a normal distribution. Therefore, the expected value μ and the standard deviation σ are calculated, and the screening rules are set according to the expected value μ and the standard deviation σ. Specifically, data outside the range of [μ-Nσ, μ+Nσ] are removed, and only data within the range of [μ-Nσ, μ+Nσ] are retained.

[0160] Where μ represents the expected value and σ represents the standard deviation. Generally, N is taken as 3, but when the precision requirements of the dispensing valve are high, N is taken as 1.

[0161] In the above embodiments, using training samples with relatively dispersed data and a large total amount of data can avoid overfitting caused by insufficient data. Using training samples with relatively concentrated data and a smaller total amount of data allows for faster training and more accurate prediction results.

[0162] Furthermore, the neural network model includes: a first neural network model and a second neural network model;

[0163] The steps for training the neural network model to obtain the prediction model include:

[0164] The lift sample is input into the first neural network model, with the lift input voltage as input and the voltage corresponding to the lift displacement as output, and the first neural network model is trained to obtain the lift prediction model;

[0165] The return sample is input into the second neural network model, with the return input voltage as input and the voltage corresponding to the return displacement as output, to train the second neural network model and obtain the return prediction model;

[0166] The lift prediction model and the return prediction model are used together as the prediction model.

[0167] In some implementations, due to the hysteresis effect, the correspondence between the input voltage and output displacement of the dispensing valve differs significantly between the lift and return processes. Training separate neural network models with sample data from the lift and return processes to obtain separate lift and return prediction models can avoid erroneous predictions and improve prediction accuracy.

[0168] Further, the specific steps of inputting the input voltage to be predicted into the prediction model to predict the voltage corresponding to the output displacement and obtain the voltage corresponding to the actual output displacement include:

[0169] Determine the movement state of the dispensing valve.

[0170] When the dispensing valve is in the lift state, the input voltage to be predicted is input into the lift prediction model to obtain the lift prediction voltage value; the lift prediction voltage value is used as the voltage corresponding to the actual output displacement.

[0171] When the dispensing valve is in the return state, the input voltage to be predicted is input into the return prediction model to obtain the return prediction voltage value; the return prediction voltage value is used as the voltage corresponding to the actual output displacement.

[0172] In some implementations, due to the hysteresis effect, the correspondence between the input voltage and output displacement of the dispensing valve differs significantly between the lift and return processes, and different models are required to predict the lift and return processes.

[0173] In the actual prediction process, the motion state of the dispensing valve is first determined, and then input into the corresponding model to obtain the voltage corresponding to the actual output displacement. Using this voltage as input to control the dispensing valve eliminates the influence of hysteresis on its control, allowing for more accurate adjustment of the valve's output displacement and ensuring precise control of the spray volume.

[0174] Example 2

[0175] This invention provides a dispensing valve input voltage correction device based on a neural network model, used for performing a dispensing valve input voltage correction method considering hysteresis effects as described in the above embodiments, comprising:

[0176] Data storage module, the data storage module being used for:

[0177] The system stores multiple sample data consisting of the input voltage and output displacement of the dispensing valve during the lift and return processes, a linear model of the dispensing valve displacement, and a neural network model.

[0178] The processing module is used for:

[0179] Obtain the corresponding relationship graph;

[0180] Multiple training samples are obtained based on the correspondence graph;

[0181] The training samples are input into the neural network model, with the input voltage as the input and the voltage corresponding to the output displacement as the output, to train the neural network model and obtain a prediction model.

[0182] Obtain the input voltage to be predicted;

[0183] Input the input voltage to be predicted into the prediction model, predict the voltage corresponding to the output displacement, and obtain the voltage corresponding to the actual output displacement;

[0184] The hysteresis displacement is calculated based on the voltage corresponding to the actual output displacement and the linear model of the dispensing valve displacement.

[0185] Specifically, multiple sample data consisting of the input voltage and output displacement of the dispensing valve during the lift and return processes were obtained through experimental measurements. Multiple corresponding output displacements were obtained by measuring the same input voltage multiple times.

[0186] The device provided in this embodiment is used to collect a large amount of sample data according to the scheme provided in the above embodiment, and to train the neural network model using the input voltage to be predicted and the voltage corresponding to the actual output displacement, so that the neural network model is fitted into a prediction model for predicting the voltage corresponding to the actual output displacement based on the input voltage to be predicted.

[0187] In the actual prediction process, the input voltage to be predicted is input into the prediction model. After prediction, the voltage corresponding to the actual output displacement is obtained. Therefore, the voltage corresponding to the actual output displacement can be directly used as the input voltage to adjust to the required output displacement. This eliminates the influence of hysteresis. By using the prediction model to correlate the input voltage to be measured with the voltage corresponding to the actual output displacement after the hysteresis effect, the input voltage controlling the dispensing valve can eliminate the influence of hysteresis and more accurately control the output displacement of the dispensing valve.

[0188] Example 3

[0189] This invention provides a computer device, comprising: a memory, a processor, and a program stored in the memory for performing input voltage correction of a dispensing valve; the program for performing input voltage correction of the dispensing valve is configured to:

[0190] A method for correcting the input voltage of a dispensing valve considering the effect of hysteresis, as described in the above embodiments, is implemented.

[0191] Specifically, a computer device includes a central processing unit (CPU), which performs various appropriate actions and processes based on programs stored in read-only memory (ROM) or loaded from memory into random access memory (RAM). RAM also stores various programs and data required for system operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0192] The following components are connected to the I / O interface: input sections including keyboards, mice, etc.; output sections including cathode ray tubes (CRTs), liquid crystal displays (LCDs), and speakers; storage sections including hard disks; and communication sections including network interface cards such as LAN cards and modems. The communication sections perform communication processing via networks such as the Internet. Drives are also connected to the I / O interface as needed. Removable media, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on the drive as needed so that computer programs read from them can be installed into the storage section as required.

[0193] Example 4

[0194] This invention provides a storage medium storing a program for measuring the hysteresis displacement of a dispensing valve; when the program for measuring the hysteresis displacement of the dispensing valve is executed, it is used to:

[0195] A method for correcting the input voltage of a dispensing valve considering the effect of hysteresis, as described in the above embodiments, is implemented.

[0196] Specifically, the present invention also provides a storage medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to implement the method as described in the above embodiments, specifically a dispensing valve input voltage correction method considering hysteresis effects as described in any one of the above embodiments.

[0197] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention is not limited to the specific combination of the above-described technical features, but also includes other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in this invention.

Claims

1. A method for correcting the input voltage of a dispensing valve considering the effect of hysteresis, characterized in that, The dispensing valve includes: a piezoelectric stack (1), a lever (2), and a firing pin assembly (3); The methods include: Obtain the correspondence graph; the correspondence graph includes multiple sample data points consisting of the input voltage and output displacement of the dispensing valve during the lift and return processes, as well as the Bode plot corresponding to the linear model of the dispensing valve displacement, including: Obtain the linear model of the dispensing valve displacement, including: The dynamic differential equations of the piezoelectric stack (1), lever (2) and striker assembly (3) are established respectively, and the differential equations of the piezoelectric stack, lever and striker assembly are obtained. Establish conversion equations between the piezoelectric stack (1) motion displacement, lever (2) rotation angle and striker assembly (3) motion displacement; Based on the differential equations of the piezoelectric stack, the lever, and the firing pin assembly, as well as the conversion equation, the dynamic differential equation of the dispensing valve is calculated. The linear displacement model of the dispensing valve is calculated based on the dynamic differential equation of the dispensing valve. The sample data is obtained, including elevation sample data and return sample data; The transfer function model and the sample data are plotted on the same coordinate system, with the input voltage as the horizontal axis and the output displacement as the vertical axis, to obtain the corresponding relationship diagram. Multiple training samples are obtained based on the correspondence diagram; the training samples include input voltage and output displacement corresponding voltage; the output displacement corresponding voltage is calculated by substituting the output displacement corresponding to the input voltage into the linear displacement model of the dispensing valve. Obtain the neural network model; The training samples are input into the neural network model, with the input voltage as the input and the voltage corresponding to the output displacement as the output, to train the neural network model and obtain a prediction model. Obtain the input voltage to be predicted; Input the input voltage to be predicted into the prediction model, predict the voltage corresponding to the output displacement, and obtain the voltage corresponding to the actual output displacement; The voltage corresponding to the actual output displacement is used as the actual input voltage of the dispensing valve.

2. The method for correcting the input voltage of a dispensing valve considering hysteresis according to claim 1, characterized in that, The steps for obtaining multiple training samples based on the correspondence graph include: S41: Randomly select a point containing an unselected sample data according to the correspondence diagram to obtain the first sample point; S42: Obtain the x-coordinate and y-coordinate of the first sample point to get the first sample input voltage and the first sample output displacement; S43: Substitute the output displacement of the first sample into the linear displacement model of the dispensing valve to calculate the voltage corresponding to the output displacement; S44: Determine the magnitude of the voltage corresponding to the output displacement and the first sample input voltage. When the voltage corresponding to the output displacement is less than the first sample input voltage, take the voltage corresponding to the output displacement and the first sample input voltage as a set of rise samples; when the voltage corresponding to the output displacement is greater than the first sample input voltage, take the voltage corresponding to the output displacement and the first sample input voltage as a set of return samples. S45: Repeat steps S41-S44 until the number of lift samples and the number of return samples both reach the set total number of samples; thus obtaining multiple lift samples and multiple return samples. S46: Use the multiple elevation samples and the multiple return samples as multiple training samples.

3. The method for correcting the input voltage of a dispensing valve considering hysteresis according to claim 2, characterized in that, The training samples are input into the neural network model. Before training the neural network model, the method further includes screening the training samples, specifically including the following steps: The values ​​of the voltages corresponding to the multiple lift displacements for each lift input voltage are statistically analyzed, and the distribution curves of the voltages corresponding to the multiple lift displacements for each lift input voltage are plotted to obtain multiple distribution curves. Calculate the expected value and standard deviation of each of the distribution curves; The screening rules are set based on the aforementioned mathematical expectation and standard deviation; The sample data is filtered according to the filtering rules. The remaining sample data after filtering will be used as training samples.

4. The method for correcting the input voltage of a dispensing valve considering hysteresis according to claim 2, characterized in that, The neural network model includes: a first neural network model and a second neural network model; The steps for training the neural network model to obtain the prediction model include: The lift sample is input into the first neural network model, with the lift input voltage as input and the voltage corresponding to the lift displacement as output, and the first neural network model is trained to obtain the lift prediction model; The return sample is input into the second neural network model, with the return input voltage as input and the return displacement corresponding voltage as output, to train the second neural network model and obtain the return prediction model; The lift prediction model and the return prediction model are used together as the prediction model.

5. The method for correcting the input voltage of a dispensing valve considering hysteresis according to claim 4, characterized in that, The specific steps for inputting the input voltage to be predicted into the prediction model, predicting the voltage corresponding to the output displacement, and obtaining the voltage corresponding to the actual output displacement include: Determine the movement state of the dispensing valve. When the dispensing valve is in the lift state, the input voltage to be predicted is input into the lift prediction model to obtain the lift prediction voltage value; the lift prediction voltage value is used as the voltage corresponding to the actual output displacement. When the dispensing valve is in the return state, the input voltage to be predicted is input into the return prediction model to obtain the return prediction voltage value; the return prediction voltage value is used as the voltage corresponding to the actual output displacement.

6. A dispensing valve input voltage correction device based on a neural network model, characterized in that, A method for correcting the input voltage of a dispensing valve considering hysteresis, as described in any one of claims 1-5, includes: Data storage module, the data storage module being used for: The system stores multiple sample data consisting of the input voltage and output displacement of the dispensing valve during the lift and return processes, a linear model of the dispensing valve displacement, and a neural network model. The processing module is used for: Obtain the corresponding relationship graph; Multiple training samples are obtained based on the correspondence graph; The training samples are input into the neural network model, with the input voltage as the input and the voltage corresponding to the output displacement as the output, to train the neural network model and obtain a prediction model. Obtain the input voltage to be predicted; Input the input voltage to be predicted into the prediction model, predict the voltage corresponding to the output displacement, and obtain the voltage corresponding to the actual output displacement; The hysteresis displacement is calculated based on the voltage corresponding to the actual output displacement and the linear model of the dispensing valve displacement.

7. A computer device, characterized in that, include: The memory, processor, and program stored in the memory for performing dispensing valve input voltage correction; the program for performing dispensing valve input voltage correction is configured to: A method for correcting the input voltage of a dispensing valve considering the effect of hysteresis, as described in any one of claims 1-5, is employed.

8. A storage medium, characterized in that, The storage medium stores a program for measuring the hysteresis displacement of a dispensing valve; when the program for measuring the hysteresis displacement of the dispensing valve is executed, it is used to: A method for correcting the input voltage of a dispensing valve considering the effect of hysteresis, as described in any one of claims 1-5, is employed.

Images