Squeezing control algorithm for strip steel after quenching and conveying and squeezing equipment applying algorithm

Abstract

The invention relates to the technical field of conveying and squeezing, in particular to a strip steel after-quenching squeezing control algorithm and conveying and squeezing equipment applying the algorithm, the algorithm comprises data acquisition and data processing, the data acquisition comprises five factors of strip steel running speed, environment humidity, environment temperature, squeezing roller use duration and water quenching liquid viscosity; performing data processing on the data acquired by the data acquisition, and during data processing, firstly performing filtering processing to remove noise and then performing multi-factor association normalization processing; according to the wringing control algorithm, under the condition that external influence factors change, unnecessary overlarge extrusion force is avoided, abrasion of the wringing roller and equipment is delayed, and the wringing conveying quality of the strip steel is improved under the condition that sufficient wringing is guaranteed through dynamic adjustment.

Description

Technical Field

[0001] The present invention relates to the technical field of conveying and squeezing, in particular to a control algorithm for squeezing a strip steel after water quenching and conveying and squeezing equipment applying the algorithm. Background Art

[0002] Conventional carbon steel strip processing involves hot-dip galvanizing and hot-coating to increase product value. During production, after the strip heats up and completes the heating process, it needs to be cooled to room temperature before entering the next production process. After water quenching, the strip surface must be squeezed dry. The effectiveness of this squeeze-out process significantly impacts the quality of downstream processing and, in turn, the quality of the final product.

[0003] During the squeezing process, if the squeezing pressure is too low, it is difficult to ensure sufficient squeezing. If the squeezing pressure is too high, it will cause increased wear of the squeezing rollers and equipment. When the strip hardness is low, it will cause pressure damage and deformation to the strip, affecting the strip quality. Therefore, appropriate control of the squeezing pressure is extremely important. The strip running speed, ambient humidity, ambient temperature, the use time of the squeezing rollers, and the viscosity of the water quenching liquid are all external factors that affect the squeezing pressure. The faster the strip runs, the more pressure needs to be increased to ensure the squeezing effect, so as to squeeze out more water stains in a shorter contact time. When the ambient humidity is high and the ambient temperature is low, the water stains remaining in the strip after squeezing are difficult to evaporate. The squeezing pressure needs to be increased to reduce the residual water stains. Conversely, when the ambient humidity is low and the ambient temperature is high, the residual water stains can be appropriately increased. The increased use time of the squeezing rollers leads to a decrease in the flatness and smoothness of the roller surface, necessitating increased pressure to ensure the effect. Because the water quenching liquid contains additives, the higher its viscosity, the more difficult it is to evaporate. In this case, the squeezing pressure also needs to be increased to reduce the residual water stains.

[0004] In traditional technologies, the squeezing pressure is usually constant. When the above-mentioned external influencing factors change, the squeezing pressure remains unchanged, which may cause the squeezing pressure to be too high or too low and cannot be accurately controlled. Summary of the Invention

[0005] The object of the present invention is to provide a strip squeezing control algorithm after water quenching and a conveying and squeezing device using the algorithm, so as to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a strip steel post-quenching squeeze-out control algorithm, the algorithm comprising the following steps:

[0007] Step 1: Data collection and data processing. The data collection includes five factors: strip running speed, ambient humidity, ambient temperature, squeeze roller usage time, and water quenching liquid viscosity.

[0008] The data collected above are processed. During data processing, filtering is first performed to remove noise and then multi-factor correlation normalization is performed. Let the original data be x, and the data after multi-factor correlation normalization be x'. After multi-factor correlation normalization, the data of each factor are in the same order of magnitude.

[0009] Step 2: Design the neural network model architecture, which is divided into an input layer, a hidden layer, and an output layer. The input layer has five neurons, corresponding to the normalized data of five influencing factors: strip running speed, ambient humidity, ambient temperature, squeezing roller usage time, and water quenching liquid viscosity. The hidden layer is set to two layers to improve the nonlinear fitting ability of the model. The number of neurons in the first hidden layer is set to 10, and the number of neurons in the second hidden layer is set to 8. The activation function of the hidden layer uses the ReLU function, and its mathematical expression is: f(x) = max(0,x). The output layer has one neuron, corresponding to the pressure value of the squeezing roller. The output value is the normalized pressure data, which is converted into the actual pressure value through denormalization in practical applications.

[0010] Step 3: Perform model training. Collect squeeze pressure data under different combinations of factors through historical production data as training samples. Use the back propagation algorithm for model training. The Adam optimizer is selected as the optimizer. The mean square error is used as the loss function. Its mathematical expression is:

[0011]

[0012] Among them, n is the number of training samples, y i is the actual output value, The model predicts the value; the training data is divided into a training set and a validation set. During the training process, the weights and biases of the neural network are continuously adjusted to minimize the loss function. At the same time, the loss value of the validation set is monitored to prevent the model from overfitting. When the loss value of the validation set no longer decreases within several consecutive training cycles, the training is stopped and the optimal model parameters are saved.

[0013] Step 4: Model prediction. When the real-time pre-processed data is obtained, it is input into the trained neural network model. After forward propagation calculation, the normalized predicted squeeze pressure value is obtained, and then it is converted into the actual pressure value through denormalization processing.

[0014] Step five, PID control adjustment. In the PID controller, the squeezing pressure value predicted by the neural network model in step four is used as the given value, and the actual pressure value of the current squeezing roller is used as the feedback value. The deviation e(t) between the given value and the feedback value is calculated. The controller output u(t) is calculated according to the PID control law, and the controller output signal is sent to the pressure actuator of the conveying squeezing equipment to adjust the pressure of the squeezing roller so that the actual pressure approaches the predicted value.

[0015] In step 1, during the multi-factor correlation normalization process, continuous operating data for at least one month during the strip squeezing process is collected, covering the strip running speed, ambient humidity, ambient temperature, squeezing roller usage time, water quenching liquid viscosity data under different operating conditions, and the corresponding squeezing roller pressure values, to form a complete data set;

[0016] Based on the data set, the Pearson correlation coefficient is used to calculate the linear correlation between the factors. The two variables are defined as X and Y. The calculation formula of the Pearson correlation coefficient r is:

[0017]

[0018] Among them, x i and y i are the i-th observation values ​​of variables X and Y, respectively. and are the means of variables X and Y, respectively, and n is the number of observations; the observations come from the specific measurements of each variable in the data set, the mean is the average value calculated from all the observations of the variable, and the number of observations is the total number of samples in the data set;

[0019] The correlation coefficients between the five factors of strip running speed, ambient humidity, ambient temperature, squeeze roller usage time, and water quenching liquid viscosity are calculated in sequence to obtain a 5×5 correlation coefficient matrix R, R ij Represents the correlation coefficient between the i-th factor and the j-th factor, where i, j = 1, 2, 3, 4, 5.

[0020] On the basis of the correlation coefficient matrix R, the correlation coefficient is quantified, the factor association matrix A is constructed, and the threshold α is set. When |R ij When |≥α, it indicates that there is a strong correlation between the two factors. ij =|R ij |;When|R ij When |<α, it is believed that the correlation between the two factors is weak. ij =0;

[0021] Through the above processing, a factor correlation matrix A containing only strongly correlated factors and their correlation degrees is obtained. This matrix can clearly reflect the mutual relationship between the factors and provide a basis for the dynamic adjustment of normalization parameters.

[0022] In the multi-factor association normalization, the association weight is introduced to improve it. The improved multi-factor association normalization formula is as follows:

[0023]

[0024] Among them, x iis the original data of the i-th factor, and are the minimum and maximum values ​​of the i-th factor, respectively, x' i is the normalized data of the i-th factor, A ij is the element in the i-th row and j-th column of the factor association matrix, indicating the degree of association between the i-th factor and the j-th factor.

[0025] In step 4, the specific formula for denormalization is: x = x' × (x max -x min )+x min ; where x' is the normalized pressure value output by the model, x max and x min It is the maximum and minimum value of pressure data in historical statistics.

[0026] A conveying and squeezing device adopts a strip steel post-quenching squeezing control algorithm to control the squeezing pressure, the conveying and squeezing device comprising:

[0027] Magnetoelectric speed sensor, used to detect the running speed of the strip;

[0028] Humidity sensor, used to detect the humidity of the working environment of the conveying and squeezing equipment;

[0029] Temperature sensor, used to detect the temperature of the working environment of the conveying and squeezing equipment;

[0030] A time counter for accumulating the usage time of the squeeze roller;

[0031] The viscosity sensor is used to detect the viscosity of the water quenching liquid during water quenching.

[0032] The conveying and squeezing device further comprises: a squeezing roller and a pressure actuator for applying squeezing pressure to the squeezing roller;

[0033] The two ends of the squeezing roller are symmetrically provided with left and right leveling devices, upper and lower leveling devices and adjustable stroke limit devices; the left and right leveling devices are used to adjust the squeezing roller left and right, the upper and lower leveling devices are used to adjust the squeezing roller up and down, and the adjustable stroke limit device is used to limit the squeezing roller.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The squeezing control algorithm of the present invention dynamically controls the pressure of the squeezing roller in real time by combining five factors: the running speed of the strip, the ambient humidity, the ambient temperature, the usage time of the squeezing roller and the viscosity of the water quenching liquid. When the above external influencing factors change, the algorithm dynamically adjusts the pressure to ensure sufficient squeezing, avoids unnecessary excessive squeezing pressure, delays the wear of the squeezing roller and equipment, and improves the quality of strip squeezing and conveying.

[0036] In the squeezing control algorithm of the present invention, the multi-factor correlation normalization is set up compared with the traditional normalization processing. It can analyze the correlation between five factors such as the strip running speed and ambient humidity, construct a factor correlation matrix, and dynamically adjust the normalization parameters according to the degree of correlation between the factors. The normalized data not only eliminates the dimensional influence, but also better reflects the influence of the synergistic effect of various factors on the squeezing roller pressure, thereby improving the accuracy and efficiency of the neural network model training.

[0037] The conveying and squeezing equipment in the present invention, through the left and right leveling devices, the upper and lower leveling devices and the adjustable stroke limit devices, enables the squeezing roller to be effectively adjusted in the left and right and up and down dimensions. Compared with the fixed structure in traditional technology, after the squeezing roller has a certain wear, it can ensure that the center line of the roller does not shift through fine-tuning, thereby ensuring the squeezing effect and reducing the probability of plate defects such as waves and warping during strip squeezing and conveying. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a structural schematic diagram of the conveying and squeezing equipment of the present invention.

[0039] Figure 2 It is a front view of the conveying and squeezing device of the present invention.

[0040] In the figure: 1. Squeezing roller; 2. Pressure actuator; 3. Left and right leveling device; 4. Up and down leveling device; 5. Adjustable stroke limit device. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] The present invention provides a technical solution: a strip steel post-quenching squeeze control algorithm, the algorithm comprising the following steps:

[0043] Step 1: Data collection and data processing. The data collection includes five factors: strip running speed, ambient humidity, ambient temperature, squeeze roller usage time, and water quenching liquid viscosity.

[0044] The data collected above is processed. During data processing, filtering is first performed to remove noise and then multi-factor correlation normalization is performed. The filtering process uses the median filtering algorithm to remove noise. Median filtering is a nonlinear filtering method that has a good inhibitory effect on impulse noise. For a sampling sequence of length N, the middle value is taken as the filter output value. The specific steps are: sort the collected N consecutive data from small to large, and take the data in the middle position as the current filtered data. In practical applications, N is usually an odd number to achieve a balance between filtering effect and real-time performance.

[0045] Assume that the original data is x, and the data after multi-factor correlation normalization processing is x'. After multi-factor correlation normalization processing, the data of each factor are at the same order of magnitude. During the multi-factor correlation normalization processing, at least one month of continuous operation data of the strip squeezing process is collected, covering the strip running speed, ambient humidity, ambient temperature, squeezing roller usage time, water quenching liquid viscosity data under different working conditions, and the corresponding squeezing roller pressure value to form a complete data set. The data collection frequency remains consistent with the original algorithm to ensure the timeliness and continuity of the data.

[0046] Based on the data set, the Pearson correlation coefficient is used to calculate the linear correlation between the factors. The two variables are defined as X and Y. The calculation formula of the Pearson correlation coefficient r is:

[0047]

[0048] Among them, x i and y i are the i-th observation values ​​of variables X and Y, respectively. and are the means of variables X and Y, respectively, and n is the number of observations; the observations come from the specific measurements of each variable in the data set, the mean is the average value calculated from all the observations of the variable, and the number of observations is the total number of samples in the data set;

[0049] The correlation coefficients between the five factors of strip running speed, ambient humidity, ambient temperature, squeeze roller usage time, and water quenching liquid viscosity are calculated in sequence to obtain a 5×5 correlation coefficient matrix R, R ij Represents the correlation coefficient between the i-th factor and the j-th factor, where i, j = 1, 2, 3, 4, 5.

[0050] On the basis of the correlation coefficient matrix R, the correlation coefficient is quantified, the factor association matrix A is constructed, and the threshold α is set. When |R ij When |≥α, it indicates that there is a strong correlation between the two factors. ij =|R ij |;When|Rij When |<α, it is believed that the correlation between the two factors is weak. ij =0;.

[0051] Through the above processing, a factor correlation matrix A containing only strongly correlated factors and their correlation degrees is obtained. This matrix can clearly reflect the mutual relationship between the factors and provide a basis for the dynamic adjustment of normalization parameters.

[0052] In the multi-factor association normalization, the association weight is introduced to improve it. The improved multi-factor association normalization formula is as follows:

[0053]

[0054] Among them, x i is the original data of the i-th factor, and are the minimum and maximum values ​​of the i-th factor, respectively, x' i is the normalized data of the i-th factor, A ij is the element in the i-th row and j-th column of the factor association matrix, indicating the degree of association between the i-th factor and the j-th factor.

[0055] During normalization, the normalization parameters of each factor are dynamically adjusted according to the factor association matrix A. The specific process is as follows:

[0056] For the strip running speed factor (assuming it is the first factor), when calculating its normalized value x'1, not only its own value range is considered It will also be combined with the degree of correlation with other factors A 1j , the value range of other factors is included in the calculation, for example, if the correlation degree between ambient humidity (the second factor) and strip running speed is A 12 Larger, then when calculating x'1, the range of ambient humidity The impact on the results will also increase accordingly.

[0057] Similarly, when calculating the normalized values ​​of other factors, such as ambient humidity, ambient temperature, usage time of the squeeze roller, and viscosity of the water quenching liquid, the improved formula is used to comprehensively consider the normalized parameter's value range and the degree of correlation with other factors to achieve dynamic adjustment of the normalized parameters. In this way, the normalized data can better reflect the impact of the synergistic effect of various factors on the squeeze roller pressure, thereby improving the validity and pertinence of the data.

[0058] Step 2: The neural network model architecture is designed, which is divided into an input layer, a hidden layer, and an output layer. The input layer has 5 neurons, which correspond to the normalized data of the five influencing factors: strip running speed, ambient humidity, ambient temperature, squeezing roller usage time, and water quenching liquid viscosity. The hidden layer is set to two layers to improve the nonlinear fitting ability of the model. The number of neurons in the first hidden layer is set to 10, and the number of neurons in the second hidden layer is set to 8. The activation function of the hidden layer adopts the ReLU function, and its mathematical expression is: f(x) = max(0,x). The ReLU function has the advantages of simple calculation and fast convergence speed, which can effectively alleviate the gradient disappearance problem and improve the training efficiency of the model.

[0059] The output layer has one neuron, corresponding to the pressure value of the squeeze roller. The output value is the normalized pressure data, which is converted to the actual pressure value through denormalization in actual application.

[0060] Step three: Model training. Historical production data is used to collect squeeze pressure data under different combinations of factors as training samples. To ensure diversity and representativeness of the training data, the collected data should cover various equipment operating conditions, including different strip speed ranges (e.g., 0-20 m / s), ambient humidity ranges (e.g., 20%-90% RH), ambient temperature ranges (e.g., 20°C-80°C), squeeze roller operating time ranges (e.g., 0-1000 hours), and water quenching solution viscosity ranges (e.g., 1-10 mPa·s). The value interval for each factor is appropriately set based on actual production conditions to ensure that the sample data covers the entire operating range.

[0061] The back propagation algorithm is used for model training. The Adam optimizer is selected as the optimizer, and the mean square error is used as the loss function. Its mathematical expression is:

[0062]

[0063] Among them, n is the number of training samples, y i is the actual output value, The training data is divided into a training set and a validation set. The training set and validation set can be divided into a ratio of 7:3. During the training process, the weights and biases of the neural network are continuously adjusted to minimize the loss function. At the same time, the loss value of the validation set is monitored to prevent overfitting of the model. When the loss value of the validation set no longer decreases within several consecutive training cycles, the training is stopped and the optimal model parameters are saved.

[0064] Step 4: Model prediction. When the real-time pre-processed data is obtained, it is input into the trained neural network model. After forward propagation calculation, the normalized predicted squeeze pressure value is obtained, and then it is converted into the actual pressure value through denormalization processing. The specific formula for denormalization processing is: x = x'×(x max -x min )+x min ; where x' is the normalized pressure value output by the model, x max and x min It is the maximum and minimum value of pressure data in historical statistics.

[0065] Step five, PID control adjustment. In the PID controller, the squeezing pressure value predicted by the neural network model in step four is used as the given value, and the actual pressure value of the current squeezing roller is used as the feedback value. The deviation e(t) between the given value and the feedback value is calculated. The controller output u(t) is calculated according to the PID control law, and the controller output signal is sent to the pressure actuator of the conveying squeezing equipment to adjust the pressure of the squeezing roller so that the actual pressure approaches the predicted value, thereby realizing real-time dynamic control of the pressure of the squeezing roller.

[0066] The PID controller is a linear controller that controls the controlled object through a linear combination of proportional (P), integral (I), and differential (D) based on the deviation between the given value and the feedback value. Its control law is:

[0067]

[0068] Where u(t) is the controller output, K p is the proportionality coefficient, K i is the integral coefficient, K d is the differential coefficient, e(t) is the deviation between the given value and the feedback value, and the three parameters K of the PID controller are p , K i , K d It needs to be adjusted according to the actual situation and will not be elaborated here.

[0069] A conveying and squeezing device adopts a squeezing control algorithm after strip quenching to control the squeezing pressure. The conveying and squeezing device includes:

[0070] Magnetoelectric speed sensor, used to detect the running speed of the strip;

[0071] Humidity sensor, used to detect the humidity of the working environment of the conveying and squeezing equipment;

[0072] Temperature sensor, used to detect the temperature of the working environment of the conveying and squeezing equipment;

[0073] A time counter for accumulating the usage time of the squeeze roller;

[0074] The viscosity sensor is used to detect the viscosity of the water quenching liquid during water quenching.

[0075] See also Figure 1 and Figure 2 The conveying and squeezing device further comprises: a squeezing roller 1 and a pressure actuator 2 for applying squeezing pressure to the squeezing roller 1. The pressure actuator 2 is a hydraulic cylinder in this embodiment;

[0076] The left and right leveling devices 3, the upper and lower leveling devices 4 and the adjustable stroke limiter 5 are symmetrically provided at both ends of the squeeze roller 1; the left and right leveling devices 3 are used to adjust the squeeze roller 1 left and right, the upper and lower leveling devices 4 are used to adjust the squeeze roller 1 up and down, and the adjustable stroke limiter 5 is used to limit the squeeze roller 1. The left and right leveling devices 3 and the upper and lower leveling devices 4 are both composed of a screw structure, and manual fine adjustment can be achieved by rotating the screw; the adjustable stroke limiter 5 is as shown in FIG. Figure 1 As shown in the figure, the roller has a screw that can be extended and fine-tuned. The limit adjustment is achieved by changing the extension position of the screw. This fine-tuning can ensure that the roller centerline does not deviate, ensuring the squeeze effect and reducing the probability of strip defects such as waves and warping during squeeze and conveying.

[0077] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Strip steel squeezing control algorithm after quenching, characterized by: The algorithm consists of the following steps: Step 1: Data collection and data processing. The data collection includes five factors: strip running speed, ambient humidity, ambient temperature, squeeze roller usage time, and water quenching liquid viscosity. The data collected above are processed. During data processing, filtering is first performed to remove noise and then multi-factor correlation normalization is performed. Let the original data collected be x, and the data after multi-factor correlation normalization be x'. After multi-factor correlation normalization, the data of each factor are in the same order of magnitude. Step 2: Design the neural network model architecture, which is divided into an input layer, a hidden layer, and an output layer. The input layer has five neurons, corresponding to the normalized data of five influencing factors: strip running speed, ambient humidity, ambient temperature, squeezing roller usage time, and water quenching liquid viscosity. The hidden layer is set to two layers to improve the nonlinear fitting ability of the model. The number of neurons in the first hidden layer is set to 10, and the number of neurons in the second hidden layer is set to 8. The activation function of the hidden layer uses the ReLU function, and its mathematical expression is: f(x) = max(0,x). The output layer has one neuron, corresponding to the pressure value of the squeezing roller. The output value is the normalized pressure data, which is converted into the actual pressure value through denormalization in practical applications. Step 3: Perform model training. Collect squeeze pressure data under different combinations of factors through historical production data as training samples. Use the back propagation algorithm for model training. The Adam optimizer is selected as the optimizer. The mean square error is used as the loss function. Its mathematical expression is: Among them, n is the number of training samples, y i is the actual output value, The model predicts the value; the training data is divided into a training set and a validation set. During the training process, the weights and biases of the neural network are continuously adjusted to minimize the loss function. At the same time, the loss value of the validation set is monitored to prevent the model from overfitting. When the loss value of the validation set no longer decreases within several consecutive training cycles, the training is stopped and the optimal model parameters are saved. Step 4: Model prediction. When the real-time pre-processed data is obtained, it is input into the trained neural network model. After forward propagation calculation, the normalized predicted squeeze pressure value is obtained, and then it is converted into the actual pressure value through denormalization processing. Step five, PID control adjustment. In the PID controller, the squeezing pressure value predicted by the neural network model in step four is used as the given value, and the actual pressure value of the current squeezing roller is used as the feedback value. The deviation e(t) between the given value and the feedback value is calculated. The controller output u(t) is calculated according to the PID control law, and the controller output signal is sent to the pressure actuator of the conveying squeezing equipment to adjust the pressure of the squeezing roller so that the actual pressure approaches the predicted value.

2. The strip squeeze control algorithm after water quenching according to claim 1, characterized in that: In step 1, during the multi-factor correlation normalization process, continuous operating data for at least one month during the strip squeezing process is collected, covering the strip running speed, ambient humidity, ambient temperature, squeezing roller usage time, water quenching liquid viscosity data under different operating conditions, and the corresponding squeezing roller pressure values, to form a complete data set; Based on the data set, the Pearson correlation coefficient is used to calculate the linear correlation between the factors. The two variables are defined as X and Y. The calculation formula of the Pearson correlation coefficient r is: Among them, x i and y i are the i-th observation values ​​of variables X and Y, respectively. and are the means of variables X and Y, respectively, and n is the number of observations; the observations come from the specific measurements of each variable in the data set, the mean is the average value calculated from all the observations of the variable, and the number of observations is the total number of samples in the data set; The correlation coefficients between the five factors of strip running speed, ambient humidity, ambient temperature, squeeze roller usage time, and water quenching liquid viscosity are calculated in sequence to obtain a 5×5 correlation coefficient matrix R, R ij Represents the correlation coefficient between the i-th factor and the j-th factor, where i, j = 1, 2, 3, 4, 5.

3. The strip squeezing control algorithm after water quenching according to claim 2, characterized in that: On the basis of the correlation coefficient matrix R, the correlation coefficient is quantified, the factor association matrix A is constructed, and the threshold α is set. When |R ij When |≥α, it indicates that there is a strong correlation between the two factors. ij =|R ij |;When|R ij When |<α, it is believed that the correlation between the two factors is weak. ij =0; Through the above processing, a factor correlation matrix A containing only strongly correlated factors and their correlation degrees is obtained. This matrix can clearly reflect the mutual relationship between the factors and provide a basis for the dynamic adjustment of normalization parameters.

4. The strip squeezing control algorithm after water quenching according to claim 3, characterized in that: In the multi-factor association normalization, the association weight is introduced to improve it. The improved multi-factor association normalization formula is as follows: Among them, x i is the original data of the i-th factor, and are the minimum and maximum values ​​of the i-th factor, respectively, x' i is the normalized data of the i-th factor, A ij is the element in the i-th row and j-th column of the factor association matrix, indicating the degree of association between the i-th factor and the j-th factor.

5. The strip squeezing control algorithm after water quenching according to claim 1 is characterized in that: In step 4, the specific formula for denormalization is: x = x' × (x max -x min )+x min ; where x' is the normalized pressure value output by the model, x max and x min It is the maximum and minimum value of pressure data in historical statistics.

6. A conveying and squeezing device, which adopts the strip steel post-quenching squeezing control algorithm according to any one of claims 1 to 5 to control the squeezing pressure, characterized in that: The conveying and squeezing equipment comprises: Magnetoelectric speed sensor, used to detect the running speed of the strip; Humidity sensor, used to detect the humidity of the working environment of the conveying and squeezing equipment; Temperature sensor, used to detect the temperature of the working environment of the conveying and squeezing equipment; A time counter for accumulating the usage time of the squeeze roller; The viscosity sensor is used to detect the viscosity of the water quenching liquid during water quenching.

7. The conveying and squeezing device according to claim 6, characterized in that Also includes: A squeeze roller and a pressure actuator for applying squeeze pressure to the squeeze roller; The two ends of the squeezing roller are symmetrically provided with left and right leveling devices, upper and lower leveling devices and adjustable stroke limit devices; the left and right leveling devices are used to adjust the squeezing roller left and right, the upper and lower leveling devices are used to adjust the squeezing roller up and down, and the adjustable stroke limit device is used to limit the squeezing roller.

Images