CN-121979302-A - Intelligent control method for water flow valve by combining mechanism and neural network

CN121979302ACN 121979302 ACN121979302 ACN 121979302ACN-121979302-A

Abstract

The invention provides an intelligent control method of a water flow valve integrating a mechanism and a neural network, and relates to the technical field of industrial automation control. The method comprises the steps of collecting flow and inlet pressure of a flow valve in different opening states of different systems, constructing and training a saturated opening prediction model, a saturated flow prediction model and an opening prediction model based on mechanism and neural network mixing based on collected data, judging whether the target flow is larger than the saturated flow or not by using the saturated flow prediction model when the change of the target flow is monitored, predicting the target opening by using the opening prediction model based on mechanism and neural network mixing based on mechanism if the target flow is smaller than or equal to the saturated flow, setting the opening of the flow valve as the predicted target opening, and directly performing fine adjustment by using a PID control mode if the difference between the current flow and the target flow is smaller than or equal to a set threshold after the flow is stable so as to achieve the target flow. Thus, the rapid and high-precision control of the cooling water flow in the continuous casting process can be realized.

Inventors

LIU XINHUA
CHEN LONG
DONG YUXI
ZHAO FAN
XIE JIANXIN

Assignees

北京科技大学

Dates

Publication Date: 20260505
Application Date: 20260105

Claims (10)

1. The intelligent control method for the water flow valve by combining the mechanism and the neural network is characterized by comprising the following steps of: s1, collecting flow and inlet pressure of a flow valve in different opening degrees under different system states, wherein the flow refers to cooling water flow; S2, constructing and training a saturated opening prediction model and a saturated flow prediction model based on the collected flow valve opening and the corresponding flow and inlet pressure; S3, constructing and training an opening prediction model based on the mixing of the mechanism and the neural network based on the opening of the flow valve and the corresponding flow and inlet pressure; S4, when the change of the target flow is monitored, judging whether the target flow is larger than the saturated flow or not by using a saturated flow prediction model, if so, judging that the flow cannot be reached and the opening of the flow valve is set to be a saturated opening, and if not, predicting the target opening by using an opening prediction model based on the mixture of a mechanism and a neural network and setting the opening of the flow valve to be a predicted target opening; S5, detecting the difference between the current flow and the target flow after the flow is stable, if the difference is larger than a set threshold, predicting the target opening by using an opening prediction model based on the mixture of the mechanism and the neural network, setting the opening of the flow valve as the target opening, and performing fine adjustment by using PID control after the flow is stable, and if the difference is smaller than or equal to the set threshold, performing fine adjustment by directly using a PID control mode to achieve the target flow.
2. The method for intelligently controlling the water flow valve by the fusion mechanism and the neural network according to claim 1, wherein the step of collecting the flow and the inlet pressure at different opening degrees of the flow valve in different system states comprises the following steps: Under different system states, the opening degree of the flow valve is changed, the flow rate and the inlet pressure of the flow valve at different opening degrees are recorded, and a data set consisting of the opening degree of the flow valve, the flow rate and the inlet pressure is obtained.
3. The method for intelligently controlling a water flow valve by combining a mechanism and a neural network according to claim 1, wherein the constructing and training a saturated opening degree prediction model and a saturated flow rate prediction model based on the collected opening degree of the flow valve and the corresponding flow rate and inlet pressure comprises: Determining saturated flow and saturated opening of the flow valve at different openings by using the residual flow gain; Training a first artificial neural network to obtain a saturated flow prediction model by using the opening, the flow and the inlet pressure of the flow valve at different opening as inputs and the saturated flow as output, wherein the trained saturated flow prediction model is used for outputting the saturated flow; And training the second artificial neural network by using the opening, the flow and the inlet pressure of the flow valve at different opening as inputs and the saturated opening as output to obtain a saturated opening prediction model, wherein the trained saturated opening prediction model is used for outputting the saturated opening.
4. The fusion mechanism and neural network intelligent control method of water flow valve according to claim 3, wherein the residual flow gain The definition is as follows: ; Wherein, the The flow rate is the flow rate when the opening of the flow valve is maximum; And The first step is respectively in the process of adjusting the opening of the flow valve step by step The value of the opening after the flow valve is regulated for the second time and the value of the flow; Is 100% open.
5. The fusion mechanism and neural network intelligent control method of the water flow valve according to claim 4, wherein the saturation opening degree H And saturation flow rate The method comprises the following steps: H ; ; ; Wherein, the As an intermediate variable, the number of the variables, For gain threshold, subscript To meet the requirements of When (1) 。
6. The intelligent control method for the water flow valve, which integrates the mechanism and the neural network according to claim 1, is characterized in that the opening prediction model based on the mixture of the mechanism and the neural network is realized by using the neural network to correct parameters of a physical formula, wherein the physical formula is a flow formula based on a Bernoulli equation corrected by the neural network: ; Wherein, the For the predicted target opening degree, For the target flow rate to be the same, Is the flow rate at the time of the maximum opening degree, Is the pressure drop at the maximum opening degree, In order for the pressure differential to be effective, 、 The optimization terms given for the coefficient correction network and the pressure correction network respectively, Is a function fitted by a neural network.
7. The intelligent control method for the water flow valve of the fusion mechanism and the neural network according to claim 6, wherein the coefficient correction network and the pressure correction network both adopt a feedforward neural network architecture and comprise an input layer, a plurality of hidden layers and an output layer.
8. The method for intelligently controlling a water flow valve by combining a mechanism and a neural network according to claim 1, wherein the predicting the target opening using an opening prediction model based on a mixture of the mechanism and the neural network comprises: the current opening, the current flow, the current inlet pressure and the target flow are taken as inputs, the target opening is taken as an output, and the target opening is predicted by using an opening prediction model based on a mixture of mechanisms and a neural network.
9. The utility model provides a water flow valve intelligent control equipment which characterized in that, water flow valve intelligent control equipment includes: A processor; A memory having stored thereon computer readable instructions which, when executed by the processor, implement the method of any of claims 1 to 8.
10. A computer readable storage medium having stored therein program code which is callable by a processor to perform the method of any one of claims 1 to 8.

Description

Intelligent control method for water flow valve by combining mechanism and neural network Technical Field The invention relates to the technical field of industrial automation control, in particular to an intelligent control method for a water flow valve by combining a mechanism and a neural network. Background Continuous casting (simply referred to as continuous casting) is a key step in metal forming. In the continuous casting process, molten metal continuously flows into a die, and is cooled and solidified under the action of cooling water, so that the control precision and efficiency of the cooling water flow directly influence the quality of a continuous casting product, and meanwhile, the control precision and efficiency are also key for realizing real-time regulation. Especially in process switching or high throughput experiments, the cooling water flow control objective often needs to be adjusted to a large extent, requiring a fast response of the process. Currently, proportional-integral-derivative (PID) controllers are mostly used for flow control. For example, chinese patent CN117884590a proposes a water flow control device and a control method in an aluminum casting system. The water supply is kept stable by monitoring the cooling water pressure and flow and by PID control. The method has better performance in the scene of small-amplitude flow fluctuation. However, the PID parameters are set empirically, and when the system is faced with a system with a large flow variation, nonlinearity, and delay feedback, there are problems of slow response speed, large overshoot, and oscillation, and the process control requirements of large-scale, rapid, and high-precision response cannot be satisfied. In recent years, a data-driven model typified by an Artificial Neural Network (ANN) has been proposed. For example, in chinese patent CN111456840B, an intelligent control method for cooling water flow of an internal combustion engine based on RBF neural network is proposed. The neural network is trained using the collected data to predict water temperature, and water flow is regulated according to the temperature difference. The artificial neural network has strong nonlinear fitting capability, but has poor generalization capability under the working condition outside the coverage range of training data, and is difficult to ensure the reliability in industrial scenes. In order to solve the above problems, it is needed to realize a flow control method that has high accuracy, fast response, physical meaning and stability. Disclosure of Invention The embodiment of the invention provides an intelligent control method for a water flow valve of a fusion mechanism and a neural network, which can rapidly control cooling water flow to a target value, thereby realizing rapid high-precision control of cooling water flow in a continuous casting process. The technical scheme is as follows: On the one hand, a water flow valve intelligent control method integrating a mechanism and a neural network is provided, the method is realized by water flow valve intelligent control equipment, and the method comprises the following steps: s1, collecting flow and inlet pressure of a flow valve in different opening degrees under different system states, wherein the flow refers to cooling water flow; S2, constructing and training a saturated opening prediction model and a saturated flow prediction model based on the collected flow valve opening and the corresponding flow and inlet pressure; S3, constructing and training an opening prediction model based on the mixing of the mechanism and the neural network based on the opening of the flow valve and the corresponding flow and inlet pressure; S4, when the change of the target flow is monitored, judging whether the target flow is larger than the saturated flow or not by using a saturated flow prediction model, if so, judging that the flow cannot be reached and the opening of the flow valve is set to be a saturated opening, and if not, predicting the target opening by using an opening prediction model based on the mixture of a mechanism and a neural network and setting the opening of the flow valve to be a predicted target opening; S5, detecting the difference between the current flow and the target flow after the flow is stable, if the difference is larger than a set threshold, predicting the target opening by using an opening prediction model based on the mixture of the mechanism and the neural network, setting the opening of the flow valve as the target opening, and performing fine adjustment by using PID control after the flow is stable, and if the difference is smaller than or equal to the set threshold, performing fine adjustment by directly using a PID control mode to achieve the target flow. Further, the collecting the flow and the inlet pressure when the flow valve has different openings in different system states includes: Under different system states, the opening degree of the fl