Rotary valve joint control method, device and equipment for feeding and conveying system

Abstract

The present invention belongs to the field of industrial control technology, and discloses a rotary valve joint control method, device and equipment for a feeding and conveying system, including: based on a preset time interval, collecting sensor data corresponding to each target item, obtaining multidimensional time series data, the target item at least includes the rotary valve opening, and the multidimensional time series data at least includes the first opening of the rotary valve; preprocessing the multidimensional time series data to obtain preprocessed time series data; inputting the preprocessed time series data into a pre-trained LSTM model to coordinate the opening of each rotary valve, and obtaining the second opening of each rotary valve output by the LSTM model; traversing the rotary valve, calculating the difference between the first opening and the second opening of the current rotary valve, and when the difference exceeds the preset threshold range, adjusting the opening of the current rotary valve. It can improve production efficiency and stability, and can respond to changes in production demand in a timely manner.

Description

Technical Field

[0001] The invention belongs to the technical field of industrial control, and in particular relates to a rotary valve combined control method, device and equipment for a feeding and conveying system. Background Art

[0002] In the polyvinyl chloride (PVC) feeding and conveying system, rotary valves are widely used to control the flow of materials to ensure that the material supply at each stage of the production process meets the process requirements. The existing rotary valve control technology mainly adopts a combination of independent control and sequence cycle. Independent control means that each rotary valve operates independently through its own sensor and controller to ensure that the material supply in a specific stage meets the requirements. Sequence cycle means that the material is transported to different powder tanks in sequence, and it is usually necessary to wait for one valve operation to be completed before starting the next valve.

[0003] However, in a complex production environment, firstly, the sequence cycle method limits the parallel operation of the rotary valve, resulting in unnecessary waiting time during the material transportation process, reducing the overall production efficiency. Secondly, although each rotary valve can accurately adjust its own material flow, due to the lack of a global coordination mechanism, it is easy to have oversupply or undersupply of materials at certain stages, which in turn affects the overall stability of the system. Thirdly, when production demand changes, traditional control methods usually rely on manual intervention, which makes it difficult to quickly respond to dynamic changes, reducing the automation level of the system. Summary of the invention

[0004] The object of the present invention is to provide a rotary valve joint control method, device, equipment and storage medium for a feeding and conveying system, which can improve the production efficiency and stability of the feeding and conveying system and can promptly respond to changes in production requirements of the feeding and conveying system.

[0005] The first aspect of the present invention discloses a rotary valve joint control method for a feeding and conveying system, which is used in a polyvinyl chloride feeding and conveying system, comprising:

[0006] Based on a preset time interval, sensor data corresponding to each target item is collected to obtain multi-dimensional time series data, wherein the target item at least includes the opening of the rotary valve, and the multi-dimensional time series data at least includes the first opening of the rotary valve;

[0007] Preprocessing the multidimensional time series data to obtain preprocessed time series data;

[0008] Inputting the preprocessed time series data into a pre-trained LSTM model to coordinate the openings of the various rotary valves, and obtaining a second opening of each rotary valve output by the LSTM model;

[0009] The rotary valve is traversed to calculate the difference between the first opening and the second opening of the current rotary valve, and when the difference exceeds a preset threshold range, the opening of the current rotary valve is adjusted.

[0010] In some embodiments, preprocessing the multidimensional time series data to obtain preprocessed time series data includes:

[0011] Filling missing values ​​in the multidimensional time series data using a spline interpolation method;

[0012] Using a moving smoothing method to replace outliers in the multidimensional time series data;

[0013] The multidimensional time series data is standardized to obtain the preprocessed time series data.

[0014] In some embodiments, the step of replacing outliers in the multidimensional time series data using a moving smoothing method includes:

[0015] According to all the data of each dimension in the multidimensional time series data, respectively calculate the mean of each dimension;

[0016] Traverse the dimensions of the multidimensional time series data, calculate the standard deviation of each data point in the multidimensional time series data in the current dimension according to the mean of the current dimension, and determine the outliers in the multidimensional time series data in the current dimension according to the standard deviation; based on the positions of the outliers in the current dimension, respectively calculate the smoothed values ​​using a moving smoothing method, and replace the outliers with the smoothed values.

[0017] In some embodiments, the loss function used when training the LSTM model is a loss function constructed based on the prediction error of the rotary valve opening, the balance of the material flow rate, the stability of the pipeline pressure, and the deviation between the production capacity and the target.

[0018] In some embodiments, the feeding and conveying system is provided with a first main pipeline and a second main pipeline, and the material flow rate balance term is used in the loss function to characterize the balance of the material flow rate, and the calculation formula of the material flow rate balance term is:

[0019] ,

[0020] in, is the material flow rate balance item, N is the total number of samples, hour, is the flow rate of the first main pipeline, hour, is the flow rate of the second main pipeline.

[0021] In some embodiments, based on the flow rate calculated in the second main pipeline, a flow loss term is used in the loss function to characterize the deviation between the production capacity and the target, and the calculation formula of the flow loss term is:

[0022] ,

[0023] in, is the flow loss term, N is the total number of samples, A is the cross-sectional area of ​​the second main pipe, is the flow rate of the second main pipe, is the actual flow rate of the second main pipeline, The target capacity.

[0024] In some embodiments, the feeding and conveying system is provided with a first fan and a second fan, and a pressure stability term is used in the loss function to characterize the stability of the pipeline pressure. The calculation formula of the pressure stability term is:

[0025]

[0026] in, is the pressure stability term, N is the total number of samples, hour, is the inlet pressure of the first fan, is the outlet pressure of the first fan; hour, is the inlet pressure of the second fan, is the outlet pressure of the second fan.

[0027] The second aspect of the present invention discloses a rotary valve combined control device for a feeding and conveying system, comprising:

[0028] A collection module, used to collect sensor data corresponding to each target item based on a preset time interval to obtain multi-dimensional time series data, wherein the target item at least includes the opening of the rotary valve, and the multi-dimensional time series data at least includes the first opening of the rotary valve;

[0029] A preprocessing module, used for preprocessing the multidimensional time series data to obtain preprocessed time series data;

[0030] A prediction module, used for inputting the preprocessed time series data into a pre-trained LSTM model to obtain a second opening of each rotary valve;

[0031] The control module is used to traverse the rotary valve, calculate the difference between the first opening and the second opening of the current rotary valve, and adjust the opening of the current rotary valve when the difference exceeds a preset threshold range.

[0032] The third aspect of the present invention discloses an electronic device, comprising a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the rotary valve joint control method for a feeding and conveying system disclosed in the first aspect.

[0033] A fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the rotary valve joint control method for a feeding and conveying system disclosed in the first aspect.

[0034] The beneficial effects of the present invention are that, by collecting multidimensional time series data through sensors, the LSTM model extracts time series features from these multidimensional time series data, predicts the opening of each rotary valve, and jointly adjusts the opening of each rotary valve to ensure the stability, efficiency and energy consumption optimization in the material transportation process, avoid blockage and insufficient supply, and respond to changes in production demand of the feeding and conveying system in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings herein show specific examples of the technical solutions described in the present invention, and together with the specific implementation methods, constitute a part of the specification, and are used to explain the technical solutions, principles and effects of the present invention.

[0036] Unless otherwise specified or defined, the same reference numerals in different drawings represent the same or similar technical features, and the same or similar technical features may also be represented by different reference numerals.

[0037] Figure 1 It is a flow chart of a rotary valve joint control method for a feeding and conveying system disclosed in an embodiment of the present invention;

[0038] Figure 2 is a flow chart of the preprocessing steps of an embodiment of the present invention;

[0039] Figure 3 is a structural schematic diagram of a rotary valve combined control device for a feeding and conveying system according to an embodiment of the present invention;

[0040] Figure 4 It is a structural schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] Unless otherwise specified or defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art. In combination with the technical solution of the present invention in a realistic scenario, all technical and scientific terms used herein may also have meanings corresponding to the purpose of implementing the technical solution of the present invention. The "first, second..." used herein is only used to distinguish the names and does not represent a specific quantity or order. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0042] It should be noted that when a component is considered to be "fixed to" another component, it can be directly fixed to the other component or there can be a central component; when an component is considered to be "connected to" another component, it can be directly connected to the other component or there can be a central component at the same time; when an component is considered to be "installed on" another component, it can be directly installed on the other component or there can be a central component at the same time. When an component is considered to be "set on" another component, it can be directly set on the other component or there can be a central component at the same time.

[0043] Unless otherwise specified or defined, the "said" and "the" used in this document refer to the technical features or technical contents mentioned or described before the corresponding position, and the technical features or technical contents may be the same as or similar to the technical features or technical contents mentioned therein. In addition, the terms "including" and "having" and any variations thereof used in this document are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0044] The feeding and conveying system in this embodiment is a polyvinyl chloride (PVC) feeding and conveying system. There are a first main pipeline and a second main pipeline in the polyvinyl chloride feeding and conveying system. Two rotary valves are installed on the first main pipeline (for the convenience of description, marked as: TV-0101, TV-0102), and four rotary valves are installed on the second main pipeline (for the convenience of description, marked as: TV-0103, TV-0104, TV-0105, TV-0106). The process flow of the polyvinyl chloride feeding and conveying system is as follows: the material is packaged in ton bags, transported to the designated ton bag unpacking equipment by a crane, and unpacked and discharged by two ton bag unpacking machines in turn, and then the unloading switch is controlled by a butterfly valve and the unloading amount is controlled by a rotary valve (TV-0101, TV-0102) to discharge the material to the accelerator. Next, the material passes through the accelerator and is conveyed by the Roots blower under dilute positive pressure, and then enters the powder tank in turn. The material is discharged to the accelerator by the rotary valve (TV-0103, TV-0104, TV-0105, TV-0106) to control the discharge amount. The material passes through the accelerator and is extracted to the vacuum feeder by the Roots blower under negative pressure, and then is conveyed to the glue liquid metering tank. In the metering tank, the material is preheated in advance, and then after the material enters the metering tank and is stirred, it is conveyed to the next process through the pipeline to complete the entire conveying process.

[0045] Due to the complexity of the design of the feeding and conveying system and the consideration of the risk of pipeline blockage, the existing PVC feeding and conveying system usually adopts an independent control and sequence cycle operation mode. For example, during the unpacking process of the ton bag, the rotary valves (TV-0101, TV-0102) of the two unpacking machines are opened in sequence: only one valve is in working state at a time, and the feeding and conveying system will switch and open the other valve after the first valve completes the unloading. Similarly, the rotary valves (TV-0103, TV-0104, TV-0105, TV-0106) of the powder tank are also opened one by one. When the liquid level of a powder tank reaches the set value, the feeding and conveying system switches to the next powder tank for filling. This operation logic is that the rotary valves in the same group work one by one, rather than opening at the same time. Although this sequential operation within the group can reduce the risk of pipeline blockage and simplify the control process, since the valves in the same group cannot work at the same time, it often leads to low production efficiency, material transportation interruption and delayed response.

[0046] In order to solve the above problems, this embodiment uses the LSTM (Long Short-Term Memory network) model to collect time series data related to the rotary valve opening, material flow rate, pipeline pressure, fan speed, etc. to predict the opening percentage of each rotary valve in the future time step, and adjust the opening of the rotary valve in real time to ensure the stability, transportation efficiency and energy consumption optimization during the material transportation process, and can respond to the dynamic changes of the feeding and conveying system in a timely manner. It should be noted that the feeding and conveying system is not limited to the polyvinyl chloride feeding and conveying system of this embodiment, and can also be used for the feeding and conveying systems of other products that require joint control of multiple valves.

[0047] The rotary valve joint control method for a feeding and conveying system disclosed in the embodiment of the present invention can be implemented by computer programming as an independent control system or as a part of the control system in a polyvinyl chloride feeding and conveying system. The execution subject of the method can be an electronic device such as a computer, a laptop, a tablet computer, or a control chip embedded in an electronic device, and the present invention is not limited to this.

[0048] In order to facilitate the understanding of the present invention, specific embodiments of the present invention will be described in more detail below with reference to the accompanying drawings.

[0049] like Figure 1 As shown, the method comprises the following steps:

[0050] Step S100: based on a preset time interval, collecting sensor data corresponding to each target item to obtain multi-dimensional time series data, wherein the target item at least includes the opening of the rotary valve, and the multi-dimensional time series data at least includes the first opening of the rotary valve.

[0051] The target items are related parameter items in the feeding and conveying system that can collect operating data and are related to the predicted opening of the rotary valve. In the PVC feeding and conveying production line of this embodiment, the target items are parameter items such as the opening, temperature, humidity, material flow rate, fan pressure, and powder tank liquid level of the rotary valve (TV-0101 to TV-0106). There is no limit to the number of target items, but at least the opening of the rotary valve must be included. Sensors can be installed in the feeding and conveying system to collect data corresponding to these target items at preset time intervals s to form multidimensional time series data. The multidimensional time series data includes at least the first opening of the rotary valve. The sensor data corresponding to the historically collected target items can be used to train the LSTM model.

[0052] Specifically, a sensor is installed at each rotary valve to obtain the real-time opening, temperature and humidity of each rotary valve; flow rate sensors are installed at the first main pipeline where TV-0101 and TV-0102 converge and the second main pipeline where TV-0103 and TV-0106 converge to collect the material flow rate; pressure sensors are installed at the air inlet and outlet of the two Roots blowers to measure the positive and negative pressure states in the feeding and transportation system in real time; liquid level sensors are installed in the four powder tanks to monitor the height of the material in the powder tanks in real time.

[0053] use represents the initial opening of the i-th rotary valve, denoted as ;use It represents the real-time opening of the i-th rotary valve, denoted as ;use represents the temperature of the i-th rotary valve, denoted as ;use represents the humidity of the i-th rotary valve, denoted as ;use , They represent the material flow rate of the first main pipeline and the material flow rate of the second main pipeline respectively, denoted as ;use , , , Indicates the pressure at the inlet and outlet of the two Roots blowers, recorded as ;use Represents the liquid level of the i-th powder tank, denoted as Multidimensional time series data is recorded as .

[0054] Step S200: preprocessing the multidimensional time series data to obtain preprocessed time series data.

[0055] During the data collection process, sensor failure or communication interruption may occur, resulting in missing data. If less data is collected at a certain moment, delete the meaningless records directly; otherwise, use the filling method to preserve the integrity of the data. Outliers may also appear due to sensor failure, measurement error, external interference, etc. If outliers are not processed, it may have an adverse effect on LSTM model training or data analysis. Therefore, it is necessary to clean the collected multidimensional time series data, process missing values, outliers, etc., and normalize or standardize the data to ensure that the numerical range of the input data is consistent.

[0056] like Figure 2 As shown, the specific steps of preprocessing in this embodiment include:

[0057] Step S210: using spline interpolation to fill missing values ​​in multidimensional time series data;

[0058] For missing values ​​in multidimensional time series data, spline interpolation is used to fill in data. The interpolation formula is as follows:

[0059] ,

[0060] Among them, t is the time point to be interpolated, , , , are the coefficients to be determined. In the interpolation process, the interpolation polynomial The following conditions need to be met: 1. At each known data point, , , that is, the data value at the two endpoints of each interval is known at that time; 2. At the boundary points of adjacent intervals At , the first derivative of the interpolation function and the second-order derivative Must be continuous, i.e. ; 3. Set boundary conditions for the entire interpolation function to constrain the behavior of the interpolation curve at the left and right ends. , , n represents the total number of known data points. Spline interpolation is a conventional technical means for data filling, which will not be described here.

[0061] Step S220: using a moving smoothing method to replace outliers in the multidimensional time series data;

[0062] For outliers in multidimensional time series data, standard deviation is first used to perform outlier detection to determine the outliers in the multidimensional time series data, and then the moving smoothing method is used to replace these outliers.

[0063] Specifically, according to all the data in each dimension of the multidimensional time series data, the mean of each dimension is calculated respectively; the dimensions of the multidimensional time series data are traversed, and for the current dimension, the standard deviation of each data point in the multidimensional time series data in the current dimension is calculated according to the mean of the current dimension, and then the outliers in the multidimensional time series data in the current dimension are determined according to the comparison result of the standard deviation with a preset threshold; according to the positions of these outliers in the current dimension, the smoothed values ​​are calculated respectively using the moving smoothing method, and then the smoothed values ​​replace the outliers.

[0064] The specific formulas for calculating the mean and standard deviation are as follows:

[0065] ,

[0066] ,

[0067] in, is the ith data point and N is the total number of data points. is the mean of the data, is the standard deviation of the data.

[0068] In this example, if the value of a data point is equal to the mean difference times, the point is considered an outlier. That is: When is an outlier.

[0069] After outliers are detected, the moving smoothing method is used to process them. The moving smoothing method replaces outliers by taking the mean or weighted mean of adjacent data points, thereby eliminating the impact of outliers on the data and making the data smoother and more continuous. The specific steps are: select a moving window size W (for example: set to 3), for each outlier data point, take W data points before and after the data point, and calculate the average of these data points. Replace the outliers with the mean, the formula is as follows:

[0070] ,

[0071] in, is the data point within the window. At the beginning and end of the data series, since there are not enough preceding and following data points for smoothing, the existing data points can be used for weighted mean.

[0072] Step S230: Standardize the multidimensional time series data to obtain preprocessed time series data.

[0073] After missing values ​​are supplemented and outliers are replaced for multidimensional time series data, standardization is performed: the data is converted to a distribution with a mean of 0 and a standard deviation of 1. The specific formula is:

[0074] .

[0075] Step S300: input the preprocessed time series data into a pre-trained LSTM model to coordinate the openings of the various rotary valves, and obtain the second opening of each rotary valve output by the LSTM model.

[0076] The LSTM model is suitable for processing input data with time dependence and can capture long-term dependencies. Therefore, this embodiment uses the LSTM model to process complex time series data and predict the optimal opening of each rotary valve in multiple future time steps, so that the feeding and conveying system can effectively respond to dynamic environmental changes and material demand fluctuations. After the pre-trained LSTM model is integrated into the PVC feeding and conveying system, the pre-processed time series data, such as Input the LSTM model, balance the rotary valve opening error, material flow rate fluctuation, pipeline pressure fluctuation and flow error in the LSTM model, coordinate each rotary valve, predict the optimal opening of each rotary valve in real time, select the optimal opening of the future time step closest to the current time step among multiple future time steps output by the LSTM model as the second opening, and obtain the second opening of each rotary valve .

[0077] Among them, the process of training the LSTM model using historical multidimensional time series data is as follows: the historical multidimensional time series data is divided into a training set and a validation set, and the time order needs to be retained when dividing the data set. The model parameters are adjusted through the training set to learn the nonlinear relationship between the rotary valve opening and various system parameters. A loss function that comprehensively considers key factors such as the rotary valve opening, pipeline pressure, and material flow rate is also introduced to avoid pipeline blockage problems that may occur when multiple valves are opened at the same time. The performance of the model is evaluated through the validation set to ensure that the model has good generalization ability and stability under different conditions. In this embodiment, the data set is divided into a training set and a test set in a ratio of 8:2 in chronological order to ensure that the model can capture time dependence and improve the accuracy and generalization ability of the prediction.

[0078] Regarding the joint control of multiple rotary valves, in the actual PVC feeding and conveying process, the simultaneous opening of multiple rotary valves may cause the main pipeline to be blocked. Therefore, the loss function of the LSTM model should not only penalize the prediction error of the rotary valve opening, but also include additional items to measure the material flow rate and pressure stability in the pipeline to ensure that the LSTM model optimizes the material flow and prevents blockage while learning the optimal rotary valve opening. In addition, in order to ensure that the feeding and conveying system maintains stable and efficient production capacity throughout the production process, a flow loss item should also be added to the loss function to penalize the difference between the actual production capacity and the target production capacity, thereby improving the system operation efficiency.

[0079] Specifically, the loss function used in training the LSTM model is a loss function based on the prediction error of the rotary valve opening, the balance of the material flow rate, the stability of the pipeline pressure, and the deviation between the production capacity and the target. The specific formula of the loss function is:

[0080] ,

[0081] in, , , , is a hyperparameter used to balance the prediction error of the rotary valve opening , Material flow rate balance item , Pressure stability term and flow loss term Impact on total losses.

[0082] Prediction error for rotary valve opening , use the mean squared error (MSE) to measure the square difference between the LSTM model prediction value and the true value, and penalize larger errors. The specific formula is:

[0083] ,

[0084] in, is the true value of the tth sample, is the predicted value of the tth sample, and N is the total number of samples.

[0085] Fluctuations in material flow rate may cause pipeline blockage. By introducing the material flow rate balance term, excessive fluctuations in material flow rate are penalized. Therefore, the material flow rate balance term is used in the loss function to characterize the balance of the material flow rate. The PVC feeding and conveying system of this embodiment is provided with a first main pipeline and a second main pipeline, and the calculation formula of the material flow rate balance term is:

[0086] ,

[0087] in, is the material flow rate balance item, N is the total number of samples, hour, is the flow rate of the first main pipeline, hour, is the flow velocity of the second main pipeline. This loss calculates the relative change of the flow velocities V1 and V2 in adjacent time steps, which is used to constrain their fluctuation range and prevent blockage in the pipeline.

[0088] Pressure fluctuations in the pipeline will affect the instability of material flow and cause pipeline blockage. Therefore, pressure stability loss is introduced to reduce the violent fluctuation of pressure in the pipeline. The PVC feeding and conveying system of this embodiment is provided with a first Roots blower and a second Roots blower, and the calculation formula of the pressure stability term is:

[0089]

[0090] in, is the pressure stability term, N is the total number of samples, hour, is the inlet pressure of the first Roots blower, is the outlet pressure of the first Roots blower; hour, is the inlet pressure of the second Roots blower, is the outlet pressure of the second Roots blower.

[0091] The flow loss term is used to measure the difference between the actual capacity of the system and the target capacity. This loss term can punish the situation where the capacity deviates from the target, ensure the maximization of material transportation efficiency, and avoid excessive or low capacity of the feeding and conveying system. After research, this embodiment found that the amount of material after the rotary valves from TV-0103 to TV-0106 is the key node that determines the material entering the accelerator and subsequent processes. Therefore, the flow rate of the second main pipeline collected by TV-0103 to TV-0106 is selected as the basis for measuring the capacity and efficiency of the entire feeding and conveying system. The calculation formula for the flow loss term is:

[0092] ,

[0093] in, is the flow loss term, N is the total number of samples, A is the cross-sectional area of ​​the second main pipe, is the flow rate of the second main pipe, is the actual flow rate of the second main pipeline, The target capacity set for the material feeding and conveying system.

[0094] AdamW is a variant of the Adam optimization algorithm, which mainly optimizes the model through weight decay. Since AdamW has the characteristics of adaptive learning rate, it can quickly adjust model parameters in a dynamic environment, reduce the risk of overfitting, and is more suitable for processing long-running industrial systems. Therefore, this embodiment uses the AdamW optimization algorithm to optimize the LSTM model, and its optimization process is divided into two parts: gradient estimation and weight decay.

[0095] When updating the gradient, similar to Adam, AdamW uses the first-order momentum and the second-order momentum to perform a weighted moving average on the gradient. Assume are the parameters of the model, and the update process of AdamW can be described as:

[0096] First, the first-order momentum estimation is performed. The specific formula is:

[0097] ,

[0098] in, is the first-order momentum at time step t, is the decay rate hyperparameter of the first-order momentum estimate, usually set to 0.9. It controls the impact of historical gradients on the first-order momentum. is the gradient at the tth time step, that is, the derivative of the model loss function with respect to the parameters.

[0099] Then the second-order momentum estimation is performed, the specific formula is:

[0100] ,

[0101] in, is the second-order momentum at time step t, which is the weighted moving average of the current squared gradient. Determines the weight of the square of the historical gradient and is usually set to 0.999.

[0102] Then make deviation correction. Since there is deviation between the first-order and second-order momentum at the initial moment, deviation correction is required. The specific formula is:

[0103]

[0104] in, for The t-th power represents the momentum decay hyperparameter As the power decays with time step t, Same reason.

[0105] Finally, the gradient update is performed, and the gradient update process of the parameters is based on the adaptive learning rate. The specific formula is:

[0106] ,

[0107] in, is the learning rate, is a small constant, usually taken as , used to prevent the denominator from being zero.

[0108] When weight decays, the weight decay term in AdamW directly acts on the model parameters themselves. The specific expression is: ,in, is the weight decay coefficient, which is used to limit the excessive growth of parameters and prevent the model from overfitting during long-term training.

[0109] Therefore, the joint control of the rotary valves is realized by introducing the LSTM deep learning model. For example, the rotary valves TV-0101 and TV-0102 can be controlled simultaneously during the unpacking process, instead of opening them one by one. Similarly, the rotary valves TV-0103, TV-0104, TV-0105, and TV-0106 of the powder tank can also work simultaneously during the filling process. This joint control significantly optimizes the continuity and efficiency of the material conveying process, and constrains the rotary valve opening, flow rate, and pressure by designing a loss function, avoiding the problem of pipeline blockage that may be caused by the simultaneous opening of multiple valves.

[0110] Step S400: traverse the rotary valves, calculate the difference between the first opening and the second opening of the current rotary valve, and when the difference exceeds a preset threshold range, adjust the opening of the current rotary valve.

[0111] After obtaining the second opening of each rotary valve output by the LSTM model, the second opening is compared with the collected first opening, and the difference between the first opening and the second opening of each rotary valve is calculated. If the difference corresponding to rotary valve A exceeds the preset threshold range, the opening of rotary valve A is adjusted to the second opening. Therefore, it is possible to make adaptive adjustments according to the changing environmental conditions, automatically correct the opening of the rotary valve, and ensure that the feeding and conveying system maintains stable and efficient operation when the environmental conditions and material requirements change. For example: Assume that the LSTM model predicts that the opening of a rotary valve (TV-0101) should be 60%, but through the real-time data feedback from the sensor, it is found that the actual opening of the valve is 50%. At this time, the error between the model prediction value (60%) and the actual value (50%) is calculated to determine whether the error exceeds the preset threshold range. If adjustment is required, the opening of TV-0101 is automatically adjusted to 60%. The LSTM model is used to perform real-time prediction and dynamic adjustment of the openings of multiple rotary valves (TV-0101 to TV-0106) to achieve intelligent control of material flow.

[0112] In some embodiments, after the LSTM model is deployed, new data can be used to perform incremental training on the LSTM model regularly to ensure that the LSTM model can adapt to long-term changes in the feeding and conveying system and improve the prediction and generalization capabilities of the LSTM model. Through real-time input of multi-dimensional time series data, online learning and adaptive control, it can self-regulate according to different production needs and environmental conditions, thereby significantly improving the automation level and production efficiency of the PVC feeding and conveying system.

[0113] In summary, this embodiment collects multi-dimensional time series data such as rotary valve opening, material flow rate, pipeline pressure, fan speed, etc. through sensors. The LSTM model extracts time series features from these multi-dimensional time series data, predicts the opening percentage of each rotary valve in the future time step, and jointly adjusts the opening of each rotary valve to ensure stability, efficiency and energy consumption optimization during material transportation, and avoid blockage and insufficient supply. Intelligent control of each rotary valve is achieved, reducing dependence on manual intervention, and significantly improving the automation level of the PVC feeding and conveying system.

[0114] like Figure 3 As shown, based on the above-mentioned rotary valve joint control method for a feeding and conveying system, an embodiment of the present invention discloses a rotary valve joint control device for a feeding and conveying system, comprising:

[0115] The acquisition module 600 is used to acquire sensor data corresponding to each target item based on a preset time interval to obtain multi-dimensional time series data, wherein the target item at least includes the opening of the rotary valve, and the multi-dimensional time series data at least includes the first opening of the rotary valve;

[0116] A preprocessing module 610 is used to preprocess the multidimensional time series data to obtain preprocessed time series data;

[0117] A prediction module 620, configured to input the preprocessed time series data into a pre-trained LSTM model to obtain a second opening of each rotary valve;

[0118] The control module 630 is used to traverse the rotary valve, calculate the difference between the first opening and the second opening of the current rotary valve, and adjust the opening of the current rotary valve when the difference exceeds a preset threshold range.

[0119] like Figure 4 As shown, an embodiment of the present invention discloses an electronic device, including a memory 401 storing executable program codes and a processor 402 coupled to the memory 401;

[0120] The processor 402 calls the executable program code stored in the memory 401 to execute the rotary valve joint control method for the feeding and conveying system described in the above embodiments.

[0121] The embodiment of the present invention further discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the rotary valve joint control method for a feeding and conveying system described in the above embodiments.

[0122] The purpose of the above embodiments is to exemplarily reproduce and deduce the technical solution of the present invention, and to fully describe the technical solution, purpose and effect of the present invention. Its purpose is to make the public understand the disclosed content of the present invention more thoroughly and comprehensively, and it does not limit the protection scope of the present invention.

[0123] The above embodiments are not exhaustive enumerations of the present invention, and there may be multiple other implementations not listed. Any replacement and improvement made without violating the concept of the present invention shall fall within the protection scope of the present invention.

Claims

1. A rotary valve joint control method for a feeding and conveying system, characterized in that: include: Based on a preset time interval, sensor data corresponding to each target item is collected to obtain multi-dimensional time series data, wherein the target item at least includes the opening of the rotary valve, and the multi-dimensional time series data at least includes the first opening of the rotary valve; Preprocessing the multidimensional time series data to obtain preprocessed time series data; Inputting the preprocessed time series data into a pre-trained LSTM model to coordinate the openings of the various rotary valves, and obtaining a second opening of each rotary valve output by the LSTM model; Traversing the rotary valve, calculating the difference between the first opening and the second opening of the current rotary valve, and when the difference exceeds a preset threshold range, adjusting the opening of the current rotary valve; The loss function used when training the LSTM model is a loss function constructed based on the prediction error of the rotary valve opening, the balance of the material flow rate, the stability of the pipeline pressure, and the deviation between the production capacity and the target; The feeding and conveying system is provided with a first main pipeline and a second main pipeline. The material flow rate balance term is used in the loss function to characterize the balance of the material flow rate. The calculation formula of the material flow rate balance term is: ， in, is the material flow rate balance item, N is the total number of samples, hour, is the flow rate of the first main pipeline, hour, is the flow rate of the second main pipeline; Based on the flow calculated in the second main pipeline, the loss function uses a flow loss term to characterize the deviation between the production capacity and the target. The calculation formula of the flow loss term is: ， in, is the flow loss term, N is the total number of samples, A is the cross-sectional area of ​​the second main pipe, is the flow rate of the second main pipe, is the actual flow rate of the second main pipeline, The target capacity; The feeding and conveying system is provided with a first fan and a second fan. The pressure stability term is used in the loss function to characterize the stability of the pipeline pressure. The calculation formula of the pressure stability term is: in, is the pressure stability term, N is the total number of samples, hour, is the inlet pressure of the first fan, is the outlet pressure of the first fan; hour, is the inlet pressure of the second fan, is the outlet pressure of the second fan.

2. The rotary valve combined control method for a feeding and conveying system according to claim 1, characterized in that: The preprocessing of the multidimensional time series data to obtain preprocessed time series data includes: Filling missing values ​​in the multidimensional time series data using a spline interpolation method; Using a moving smoothing method to replace outliers in the multidimensional time series data; The multidimensional time series data is standardized to obtain the preprocessed time series data.

3. The rotary valve combined control method for a feeding and conveying system according to claim 2, characterized in that: The step of replacing abnormal values ​​in the multidimensional time series data by a moving smoothing method includes: According to all the data of each dimension in the multidimensional time series data, respectively calculate the mean of each dimension; Traverse the dimensions of the multidimensional time series data, calculate the standard deviation of each data point in the multidimensional time series data in the current dimension according to the mean of the current dimension, and determine the outliers in the multidimensional time series data in the current dimension according to the standard deviation; based on the positions of the outliers in the current dimension, respectively calculate the smoothed values ​​using a moving smoothing method, and replace the outliers with the smoothed values.

4. A rotary valve combined control device for a feeding and conveying system, characterized in that: include: A collection module, used to collect sensor data corresponding to each target item based on a preset time interval to obtain multi-dimensional time series data, wherein the target item at least includes the opening of the rotary valve, and the multi-dimensional time series data at least includes the first opening of the rotary valve; A preprocessing module, used for preprocessing the multidimensional time series data to obtain preprocessed time series data; A prediction module, used for inputting the preprocessed time series data into a pre-trained LSTM model to obtain a second opening of each rotary valve; A control module, configured to traverse the rotary valve, calculate the difference between the first opening and the second opening of the current rotary valve, and adjust the opening of the current rotary valve when the difference exceeds a preset threshold range; The loss function used when training the LSTM model is a loss function constructed based on the prediction error of the rotary valve opening, the balance of the material flow rate, the stability of the pipeline pressure, and the deviation between the production capacity and the target; The feeding and conveying system is provided with a first main pipeline and a second main pipeline. The material flow rate balance term is used in the loss function to characterize the balance of the material flow rate. The calculation formula of the material flow rate balance term is: ， in, is the material flow rate balance item, N is the total number of samples, hour, is the flow rate of the first main pipeline, hour, is the flow rate of the second main pipeline; Based on the flow calculated in the second main pipeline, the loss function uses a flow loss term to characterize the deviation between the production capacity and the target. The calculation formula of the flow loss term is: ， in, is the flow loss term, N is the total number of samples, A is the cross-sectional area of ​​the second main pipe, is the flow rate of the second main pipe, is the actual flow rate of the second main pipeline, The target capacity; The feeding and conveying system is provided with a first fan and a second fan. The pressure stability term is used in the loss function to characterize the stability of the pipeline pressure. The calculation formula of the pressure stability term is: in, is the pressure stability term, N is the total number of samples, hour, is the inlet pressure of the first fan, is the outlet pressure of the first fan; hour, is the inlet pressure of the second fan, is the outlet pressure of the second fan.

5. An electronic device, characterized in that: It comprises a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the rotary valve joint control method for a feeding and conveying system as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the rotary valve joint control method for a material feeding and conveying system according to any one of claims 1 to 3.

Images