PID (Proportion Integration Differentiation)-based flow regulating valve adaptive control method and system

Abstract

The invention relates to the technical field of flow control, and provides a flow regulating valve self-adaptive control method and system based on PID, and the method comprises the steps: obtaining construction parameters of all pipelines of a semiconductor, collecting pressure data and flow at different collection moments, and building a pressure sequence; marking a target sampling point and a to-be-analyzed sampling point, calculating a first gas pressure characteristic value of the target sampling point and a loss characteristic value of the to-be-analyzed sampling point, and calculating a second gas pressure characteristic value and an adjustment factor of the target sampling point; the method comprises the following steps: acquiring pressure data of a target sampling point, clustering all acquisition moments of the target sampling point to obtain a clustering cluster, determining a reference value of a PID parameter of the clustering cluster according to pressure data and flow at all acquisition moments in the clustering cluster, adjusting the reference value according to adjusting factors at different acquisition moments in the clustering cluster, and realizing adaptive control of a flow adjusting valve according to an adjusting result. The accuracy and stability of gas flow control can be improved.

Description

Technical Field

[0001] This invention relates to the field of flow control technology, and more specifically to a PID-based adaptive control method and system for flow regulating valves. Background Technology

[0002] In semiconductor manufacturing processes, using gas flow controllers to precisely regulate the flow rates of reactive gases in real time can effectively reduce fluctuations and interference during the process, ensuring product quality and consistency. For example, in CVD (Chemical Vapor Deposition) processes, the flow rates of reactive gases such as silane and oxygen must be precisely controlled at the optimal ratio to generate high-quality, uniform thin films that provide a reliable insulating layer for semiconductor devices. Inaccurate gas flow rates can lead to uneven film thickness and imbalanced composition ratios, affecting the performance and reliability of semiconductor devices and reducing product yield.

[0003] Generally, the FCM-PID algorithm, which combines FCM fuzzy C-means clustering and PID control, can be used to regulate the flow rate of process gases in semiconductor manufacturing. However, the paths of process gases to different process equipment vary under different process flows, and common process gases have dynamic and nonlinear physical characteristics. The FCM-PID algorithm is prone to being overly sensitive when faced with complex data, which often leads to insufficient stability and accuracy of gas flow control in flow regulating valves. Summary of the Invention

[0004] This invention provides a PID-based adaptive control method and system for flow control valves to address the problem that the FCM-PID algorithm is too sensitive to dynamic and nonlinear changes in process gases, resulting in insufficient stability and accuracy of gas flow control. The specific technical solution adopted is as follows: In a first aspect, one embodiment of the present invention provides an adaptive control method for a flow regulating valve based on PID control, the method comprising the following steps: Obtain the construction parameters of all semiconductor pipelines, set different sampling points in different pipeline branches, collect pressure data at different sampling times at the sampling points, establish the pressure sequence of the sampling points at the sampling times, and collect the flow rate of the process equipment at the end of each pipeline branch at each sampling time. The last sampling point of any pipeline branch is designated as the target sampling point. Based on the changing trend and relative dispersion of the pressure data of the target sampling point at adjacent sampling times, the first gas pressure characteristic value of the target sampling point at the sampling time is calculated. Any sampling point in all pipeline branches is designated as the sampling point to be analyzed. The energy of the sampling point to be analyzed at the sampling time is calculated. Based on the difference in energy and pressure sequence between the sampling point to be analyzed and its adjacent sampling points at the sampling time, the loss characteristic value of the sampling point to be analyzed at the sampling time is calculated. Based on the loss characteristic values ​​of all sampling points in the pipeline branch where the target sampling point is located at different sampling times, the second gas pressure characteristic value of the target sampling point at the sampling time is calculated. Combining the construction parameters of all pipelines and the first gas pressure characteristic value, the adjustment factor of the target sampling point at the sampling time is calculated. Based on the relative dispersion of the target sampling points at different acquisition times and the adjustment factor, clusters are obtained for all acquisition times of the target sampling points. Based on the pressure data and flow rate of all acquisition times within the same cluster, the baseline value of the PID parameter of the same cluster is determined. The baseline value is adjusted according to the difference between the adjustment factors of different acquisition times within the cluster determined by the target sampling points. The adaptive control of the flow regulating valve is realized based on the adjustment result.

[0005] Furthermore, the specific method for determining the first gas pressure characteristic value of the target sampling point at the time of sampling is as follows: Any acquisition time is designated as the target acquisition time. Based on the changing trend of pressure data at the target sampling point at the acquisition time adjacent to the target acquisition time, the pressure prediction sequence of the target sampling point at the next adjacent acquisition time is determined. The DTW distance between the pressure prediction sequence of the target sampling point at the next adjacent acquisition time and the pressure prediction sequence of the target acquisition time is designated as the pressure prediction distance of the target sampling point at the target acquisition time. The first preset number of adjacent acquisition times before the target acquisition time are recorded as the previous adjacent times of the target acquisition time. The coefficient of variation of all pressure data of the target sampling point at the target acquisition time and all previous adjacent times of the target acquisition time is recorded as the relative dispersion of the target sampling point at the target acquisition time. The positive correlation between the pressure prediction distance and relative dispersion of the target sampling point at the target acquisition time is recorded as the first gas pressure characteristic value of the target sampling point at the target acquisition time.

[0006] Furthermore, the specific method for determining the relative dispersion is as follows: The coefficient of variation of all pressure data at the target sampling point at the target acquisition time and at all the time immediately preceding the target acquisition time is denoted as the relative dispersion of the target sampling point at the target acquisition time.

[0007] Furthermore, the specific method for obtaining the loss characteristic value of the sampling point to be analyzed at the acquisition time is as follows: The sampling point adjacent to the sampling point in the pipeline branch where the sampling point to be analyzed is located is recorded as the previous adjacent sampling point. The absolute value of the energy difference between the sampling point to be analyzed and the previous adjacent sampling point at the time of sampling is recorded as the energy difference of the sampling point to be analyzed at the time of sampling. The ratio of the energy difference of the sampling point to be analyzed at the time of sampling to the energy of the previous adjacent sampling point at the time of sampling is recorded as the energy relative difference of the sampling point to be analyzed at the time of sampling. The DTW distance between the normalized pressure sequence of the sampling point to be analyzed and the previous adjacent sampling point at the acquisition time is denoted as the pressure loss difference of the sampling point to be analyzed at the acquisition time. The positive correlation between the relative energy difference and the pressure loss difference of the sampling point to be analyzed at the time of acquisition is recorded as the loss characteristic value of the sampling point to be analyzed at the time of acquisition.

[0008] Furthermore, the specific calculation method for the second gas pressure characteristic value of the target sampling point at the time of sampling is as follows: Based on the loss characteristic values ​​of all sampling points in the same pipeline branch at the same acquisition time, establish the column determined by the last sampling point in the same pipeline branch at the same acquisition time. Based on the columns determined by the last sampling point in the same pipeline branch at the target acquisition time and all the previous adjacent times of the target acquisition time, establish the loss characteristic value matrix of the same sampling point at the same acquisition time. Perform singular value decomposition on the loss eigenvalue matrix, and record the maximum value of all singular values ​​as the second gas pressure eigenvalue of the last sampling point in the same pipeline branch at the target sampling time.

[0009] Furthermore, the formula for calculating the adjustment factor of the target sampling point at the acquisition time is: in, Indicates the first The airflow obstruction of each duct branch; construction parameters include duct length and duct diameter. Indicates the first The first branch of the pipeline The length of each branch pipe; Indicates the first The sum of the lengths of all branch pipes of a given pipeline; Indicates the first The diameter of the main pipeline of each branch pipeline; Indicates the first The first branch of the pipeline The diameter of each branch pipe; Indicates the first The number of all branch pipes included in each pipe branch; Indicates the first The adjustment factor at the time of acquisition for the last sampling point of each pipeline branch; This indicates the preset weighting coefficients; Indicates the first The first gas pressure characteristic value at the time of sampling of the last sampling point of each pipeline branch; Indicates the first The second gas pressure characteristic value at the time of sampling of the last sampling point of each pipeline branch; This represents the tanh function.

[0010] Furthermore, the specific method for clustering the target sampling points across all sampling times based on the relative dispersion of the target sampling points at different sampling times and the adjustment factor includes: The adjusted metric distance between the target sampling point and the cluster center is used as the clustering metric distance in the FCM algorithm. Clustering is performed on the target sampling point at all sampling times. The formula for calculating the adjusted metric distance between the target sampling point and the cluster center is: in, Indicates the target sampling point at the time of acquisition. Collection time corresponding to cluster center Adjust the distance measurement; This indicates the sampling time corresponding to the cluster center of the target sampling point. The relative dispersion of the pressure sequence; Indicates the target sampling point at the time of acquisition. The relative dispersion of the pressure sequence; Indicates the target sampling point at the time of acquisition. The adjustment factor; This indicates the sampling time corresponding to the cluster center of the target sampling point. The adjustment factor.

[0011] Furthermore, the specific method for obtaining the baseline values ​​of the PID parameters for the same cluster is as follows: The pressure data at all sampling times within the same cluster determined by the target sampling point are fitted with the flow curves of the process equipment corresponding to the target sampling point at all sampling times within the cluster to obtain the numerical relationship between the pressure data and flow of the process equipment corresponding to the target sampling point. Combined with the PID parameter tuning method, the baseline value of the PID parameter of the same cluster is obtained.

[0012] Furthermore, the method for adjusting the baseline value based on the difference in adjustment factors between different sampling times within the cluster determined by the target sampling point, and realizing adaptive control of the flow regulating valve based on the adjustment result, includes the following specific methods: The sum of the absolute value of the difference between the adjustment factor of the target sampling point at the acquisition time and the adjustment factor at the acquisition time corresponding to the cluster center of the cluster where the acquisition time is located, and the number 1, is recorded as the adjustment range of the target sampling point at the acquisition time. The product of the adjustment range of the target sampling point at the acquisition time and the baseline value of the PID parameter of the cluster where the acquisition time is located is recorded as the adjustment value of the PID parameter of the target sampling point at the acquisition time. The adjustment value of the PID parameter at the target sampling point at the acquisition time is used as the value of the PID parameter to achieve adaptive control of the flow control valve of the process equipment corresponding to the target sampling point.

[0013] Secondly, embodiments of the present invention also provide a PID-based adaptive control system for a flow regulating valve, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0014] The beneficial effects of this invention are: Considering that the pressure of the process gas decreases as it flows from the main pipeline to each branch pipeline, and that the flow rate and pressure changes differ for branch pipelines of different diameters, the stability of pressure in each pipeline branch is evaluated during continuous sampling. The first and second gas pressure characteristic values ​​of the target sampling points at each sampling time are obtained. Combined with the construction parameters of all pipelines, the pressure loss of the process gas as it passes through each pipeline branch is further analyzed. The adjustment factor of the last sampling point in each pipeline branch at each sampling time is obtained. The greater the airflow obstruction in a pipeline branch, the longer the length of each branch pipeline. Furthermore, the relationship between the branch pipeline and the main pipeline... The greater the difference in pipe diameters, the greater the resistance encountered by the process gas as it passes through the pipe branch, and the greater the pressure change when the process gas reaches the process equipment. Furthermore, based on the relative dispersion of the target sampling points at different acquisition times and the adjustment factor, all acquisition times of the target sampling points are clustered into different clusters. The adjustment value of the PID parameter is determined based on the pressure data and flow rate of all acquisition times within the same cluster, as well as the differences in the adjustment factor between different acquisition times. This achieves adaptive control of the flow control valve, solving the problem that the FCM-PID algorithm is too sensitive to the dynamic and nonlinear changes of process gas, resulting in insufficient stability and accuracy of gas flow control, and improving the accuracy of the flow control valve's adaptive control of process gas flow. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic flowchart of a PID-based adaptive control method for a flow regulating valve provided in one embodiment of the present invention. Figure 2 This is a schematic diagram of the pipeline distribution provided in one embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 The diagram illustrates a flowchart of a PID-based adaptive control method for a flow regulating valve according to an embodiment of the present invention. The method includes the following steps: Step S001: Obtain the construction parameters of all semiconductor pipelines, set different sampling points in different pipeline branches, collect pressure data at different sampling times at the sampling points, establish the pressure sequence of the sampling points at the sampling times, and collect the flow rate of the process equipment at the end of each pipeline branch at each sampling time.

[0019] In semiconductor manufacturing, the distribution of pipelines is complex. Process gases flow in from the main pipeline, then through branch pipelines to various production equipment. Sampling points are set after valves at each branch point, and pressure sensors are installed at each sampling point to collect pressure data. Any given sampling time is designated as the target sampling time. The first predetermined number of adjacent sampling times preceding the target sampling time are designated as the previous adjacent times. The pressure data from the target sampling time and all its previous adjacent times are arranged in the order of sampling to obtain the pressure sequence for the target sampling time.

[0020] In this embodiment, the pressure data acquisition interval is set to 1 second. In practical applications, as other implementation methods, the implementer can decide the value of the pressure data acquisition interval according to the actual situation, and this application does not impose any special restrictions. In this embodiment, the first preset quantity is set to 10.

[0021] In particular, when collecting pressure data, because it takes a certain amount of time for the process gas to flow from the main pipeline to the process equipment, the pressure data collection and analysis should only begin when each process equipment is operating normally.

[0022] Pipeline distribution diagram as follows Figure 2 As shown, in Figure 2 In the diagram, black lines represent pipelines, including main pipelines and branch pipelines. A, B1, B2, C1, and C2 are the locations of sampling points. Figure 2 The branch pipelines include branch pipeline a1, branch pipeline a2, branch pipeline b1, and branch pipeline b2. Process gas enters the pipeline from the inlet and passes sequentially through the main pipeline and the branch pipelines. Figure 2 The process flow diagram includes three pipeline branches. The first branch is a main pipeline leading to branch pipeline a1, branch pipeline b1, and then to process equipment 1. The sampling points for this branch are A, B1, and C1. The second branch is also a main pipeline leading to branch pipeline a1, branch pipeline b2, and then to process equipment 2. The sampling points for this branch are A, B1, and C2. The third branch is a main pipeline leading to branch pipeline a2 and then to process equipment 3. The sampling points for this branch are A-B2. It is understood that the equipment requiring process gas flow regulation based on the FCM-PID algorithm consists of process equipment 1, process equipment 2, and process equipment 3. Process equipment 1 and 2 are different process equipment within process flow 1 corresponding to the first branch, while process equipment 3 is a process equipment within process flow 2 corresponding to the second branch. It is important to note that branch pipelines a1 and a2 have different diameters, while branch pipelines b1 and b2 have the same diameter.

[0023] In this embodiment, bilateral filtering is used to denoise the pressure data. Bilateral filtering is a well-known technique and will not be described in detail here. As other embodiments, while achieving the goal of denoising pressure data, implementers may use other methods in the prior art, such as mean filtering, for data denoising. This application does not impose any special restrictions.

[0024] The flow rate of the process equipment is collected at each sampling moment by a flow meter located at the end of each pipeline branch.

[0025] Based on the construction records, extract all construction parameters for the pipelines, including the main pipelines and branch pipelines, and the construction parameters include the length and diameter of the pipelines.

[0026] At this point, the construction parameters of all pipelines, the flow rate of the process equipment at the end of each pipeline branch at each acquisition time, and the pressure data and pressure sequence of different sampling points at different acquisition times are obtained.

[0027] Step S002: Record the last sampling point of any pipeline branch as the target sampling point. Based on the changing trend and relative dispersion of the pressure data of the target sampling point at adjacent acquisition times, calculate the first gas pressure characteristic value of the target sampling point at the acquisition time. Record any sampling point in all pipeline branches as the sampling point to be analyzed. Calculate the energy of the sampling point to be analyzed at the acquisition time. Based on the difference in energy and pressure sequence between the sampling point to be analyzed and its adjacent sampling points at the acquisition time, calculate the loss characteristic value of the sampling point to be analyzed at the acquisition time. Based on the loss characteristic values ​​of all sampling points in the pipeline branch where the target sampling point is located at different acquisition times, calculate the second gas pressure characteristic value of the target sampling point at the acquisition time. Combining the construction parameters of all pipelines and the first gas pressure characteristic value, calculate the adjustment factor of the target sampling point at the acquisition time.

[0028] From a cost perspective, the diameter of the branch pipes is smaller than that of the main pipe. According to the law of conservation of mass, the sum of the gas flow rates of all branch pipes is equal to the gas flow rate of the main pipe. At the same time, the smaller diameter pipes will bring greater flow resistance losses to the process gas. Therefore, when the process gas flows from the main pipe to each branch pipe, the pressure of the process gas in the branch pipe will decrease. For branch pipes with different diameters, the changes in flow rate and pressure in the branch pipes will also be different.

[0029] For example, when process gas flows from the main pipeline to branch pipelines a1 and a2 with different diameters, the diameters of branch pipelines a1 and a2 are both smaller than the diameter of the main pipeline. The sum of the process gas flow rates in branch pipelines a1 and a2 equals the process gas flow rate in the main pipeline. However, the process gas flow rates in branch pipelines a1 and a2 are different. If the diameter of branch pipeline a2 is smaller than the diameter of branch pipeline a1, the impedance of the process gas flowing into branch pipeline a2 is greater than the impedance of the process gas flowing into branch pipeline a1. Therefore, the process gas flow rate in branch pipeline a2 is less than the process gas flow rate in branch pipeline a1. At the same time, the pressure of the process gas in branch pipeline a2 drops faster, while the pressure drop in branch pipeline a1 is more gradual. Therefore, the pressure data of the process gas reaching different process equipment after traveling through paths of different diameters and lengths may differ. For example, the pressure data collected at sampling points C1 and B2 corresponding to process equipment 1 and process equipment 3 at the same sampling time may differ.

[0030] The last sampling point of any pipeline branch is designated as the target sampling point. Based on the pressure data change trend of the target sampling point at the next adjacent sampling time, the pressure prediction sequence of the target sampling point at the next adjacent sampling time is determined.

[0031] Using the pressure data of the target sampling point at the target acquisition time and all its preceding adjacent times as dependent variables, and the order of the acquisition times corresponding to the pressure data as dependent variables, a gradient descent algorithm is used for linear fitting to obtain the slope of the fitted line, denoted as the first slope of the target sampling point at the target acquisition time. The kurtosis and mean of the pressure data of the target sampling point at the target acquisition time and all its preceding adjacent times are calculated and denoted as the first kurtosis and first mean of the target sampling point at the target acquisition time. The vector composed of the first slope, first kurtosis, and first mean of the target sampling point at the target acquisition time is used as the feature vector. A decision tree algorithm is used to process the pressure sequence of the target sampling point at the target acquisition time and all its preceding adjacent times to obtain the pressure prediction sequence of the next adjacent acquisition time of the target sampling point.

[0032] In this embodiment, the number of iterations is set to 20 and the learning rate is set to 0.01 when using the gradient descent algorithm; when using the decision tree algorithm, the leaf node parameter is set to 2 and the maximum depth parameter is set to 5. The use of the gradient descent algorithm for linear fitting, calculation of the kurtosis and mean of the data, and the use of the decision tree algorithm are all well-known techniques and will not be described in detail here.

[0033] Based on the difference between the pressure prediction sequences and the relative dispersion of pressure data at the target sampling point at sampling times adjacent to the target sampling time, the first gas pressure characteristic value of the target sampling point at the target sampling time is calculated.

[0034] The DTW distance between the pressure prediction sequence of the target sampling point at the next adjacent sampling time and the pressure prediction sequence at the target sampling time is denoted as the pressure prediction distance of the target sampling point at the target sampling time. The coefficient of variation of all pressure data at the target sampling point at the target sampling time and all its preceding adjacent times is denoted as the relative dispersion of the target sampling point at the target sampling time. The positive correlation result between the pressure prediction distance and the relative dispersion of the target sampling point at the target sampling time is denoted as the first gas pressure characteristic value of the target sampling point at the target sampling time. The calculation of the coefficient of variation is a well-known technique and will not be elaborated further.

[0035] It is understood that a positive correlation is applied to the pressure prediction distance and relative dispersion, ensuring that the pressure prediction distance and relative dispersion are positively correlated with the first gas pressure characteristic value. It is also understood that the positive correlation in this application refers to the relationship between the independent and dependent variables, where the independent variables are the pressure prediction distance and relative dispersion, and the dependent variable is the first gas pressure characteristic value. The positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and can be an additive or multiplicative relationship.

[0036] Preferably, as an embodiment of this application, the product of the pressure prediction distance of the target sampling point at the target acquisition time and the relative dispersion is recorded as the first gas pressure characteristic value of the target sampling point at the target acquisition time.

[0037] Record any sampling point in all pipeline branches as the sampling point to be analyzed. Convert the pressure sequence of the sampling point to be analyzed at the target acquisition time to the frequency domain. Record the sum of squares of all amplitudes in the frequency domain as the energy of the sampling point to be analyzed at the target acquisition time.

[0038] The energy at each sampling point in all pipeline branches at each acquisition time can be obtained using the same method. The frequency domain transformation algorithm can utilize Fourier transform, wavelet transform, etc.

[0039] The preceding adjacent sampling point of the sampling point in the pipeline branch to be analyzed is denoted as the preceding adjacent sampling point. The absolute value of the energy difference between the sampling point to be analyzed and the preceding adjacent sampling point at the target acquisition time is denoted as the energy difference of the sampling point to be analyzed at the target acquisition time. The ratio of the energy difference of the sampling point to be analyzed at the target acquisition time to the energy of the preceding adjacent sampling point at the target acquisition time is denoted as the relative energy difference of the sampling point to be analyzed at the target acquisition time. The DTW distance of the normalized pressure sequence between the sampling point to be analyzed and the preceding adjacent sampling point at the target acquisition time is denoted as the pressure loss difference of the sampling point to be analyzed at the target acquisition time. The positive correlation result between the relative energy difference and the pressure loss difference of the sampling point to be analyzed at the target acquisition time is denoted as the loss characteristic value of the sampling point to be analyzed at the target acquisition time.

[0040] Preferably, as an embodiment of this application, the product of the relative energy difference and pressure loss difference of the sampling point to be analyzed at the target acquisition time is recorded as the loss characteristic value of the sampling point to be analyzed at the target acquisition time. In this embodiment, the sigmoid function is used to calculate the normalized value. The sigmoid function is a well-known technique and will not be described in detail here. As other implementation methods, the implementer can use other methods of the prior art, such as the tanh function.

[0041] Specifically, the first sampling point of each pipeline branch has no preceding adjacent sampling point. The sum of the maximum value of the loss characteristic value of all sampling points of all pipeline branches at the target acquisition time and a second preset number is used as the loss characteristic value of the first sampling point of all pipeline branches at the target acquisition time. In this embodiment, the second preset number is set to 0.1.

[0042] The loss characteristic values ​​of all sampling points in the same pipeline branch at the same acquisition time are arranged sequentially according to the order in which the sampling points appear in the pipeline branch. This arrangement forms the column determined by the last sampling point in the same pipeline branch at the same acquisition time. The columns determined by the last sampling point in the same pipeline branch at the target acquisition time and all its preceding adjacent times are then arranged sequentially from left to right according to the acquisition time. This creates a loss characteristic value matrix for the same sampling point at the same acquisition time. Singular value decomposition is performed on the loss characteristic value matrix, and the maximum value of all singular values ​​is recorded as the second gas pressure characteristic value of the last sampling point in the same pipeline branch at the target acquisition time.

[0043] Based on the construction parameters of all pipelines, the first gas pressure characteristic value and the second gas pressure characteristic value of the target sampling point at the target acquisition time, the adjustment factor of the target sampling point at the target acquisition time is calculated. The formula for calculating the adjustment factor is: in, Indicates the first The degree of airflow obstruction in each branch of the pipeline; Indicates the first The first branch of the pipeline The length of each branch pipe; Indicates the first The sum of the lengths of all branch pipes of a given pipeline; Indicates the first The diameter of the main pipeline of each branch pipeline; Indicates the first The first branch of the pipeline The diameter of each branch pipe; Indicates the first The number of all branch pipes included in each pipe branch; Indicates the first The adjustment factor of the last sampling point of each pipeline branch at the target acquisition time; This represents a preset weighting coefficient. The value of the weighting coefficient should be greater than 0 and less than 1. In this embodiment, the weighting coefficient is set to 0.5. Indicates the first The first gas pressure characteristic value at the target acquisition time of the last sampling point of each pipeline branch; Indicates the first The second gas pressure characteristic value at the target acquisition time of the last sampling point of each pipeline branch; This represents the tanh function, which normalizes the values ​​of the adjustment factor.

[0044] When the first gas pressure characteristic value of the last sampling point of the pipeline branch is smaller and the second gas pressure characteristic value is larger at the target acquisition time, the usage of process gas at the last sampling point of the pipeline branch at the target acquisition time is more stable, and there is less need to adjust the flow rate of process gas. At this time, the adjustment factor of the last sampling point of the pipeline branch at the target acquisition time is larger.

[0045] The longer the length of each branch pipe in a pipeline branch, and the greater the difference in diameter between the branch pipe and the main pipe, the greater the resistance encountered by the process gas as it passes through the pipeline branch, and the greater the pressure change when the process gas reaches the process equipment. In this case, the airflow resistance of the pipeline branch is greater.

[0046] The same method can be used to obtain the adjustment factor of the last sampling point in each pipeline branch at each sampling time.

[0047] At this point, the adjustment factor of the last sampling point in each pipeline branch at each sampling time is obtained.

[0048] Step S003: Based on the relative dispersion of the target sampling point at different acquisition times and the adjustment factor, cluster all acquisition times of the target sampling point to obtain clusters. Based on the pressure data and flow rate of all acquisition times within the same cluster, determine the reference value of the PID parameter of the same cluster. Adjust the reference value according to the difference between the adjustment factors of different acquisition times within the cluster determined by the target sampling point. Based on the adjustment result, realize the adaptive control of the flow regulating valve.

[0049] Based on the relative dispersion and adjustment factor of the sampling points at different acquisition times, the adjusted metric distance between each acquisition time corresponding to the sampling point and the acquisition time corresponding to the cluster center is calculated. The adjusted metric distance between the target sampling point and the cluster center is used as the clustering metric distance in the FCM algorithm. The target sampling points are clustered at all acquisition times to obtain clusters.

[0050] In this example, the FCM algorithm, or Fuzzy C-Means algorithm, is used for clustering. The number of clusters is set to 10, the fuzziness factor parameter to 2, and the number of iterations to 100. It's important to note that to ensure the clustering is meaningful, starting from the [number missing] iteration... Analysis begins from each data collection point. This indicates a preset quantity; in this embodiment, it refers to... The value is 20, that is The value is the sum of the number of clusters and the first preset number.

[0051] The formula for calculating the adjusted distance between the target sampling point and the cluster center is: in, Indicates the target sampling point at the time of acquisition. Collection time corresponding to cluster center Adjust the distance measurement; This indicates the sampling time corresponding to the cluster center of the target sampling point. The relative dispersion of the pressure sequence; Indicates the target sampling point at the time of acquisition. The relative dispersion of the pressure sequence; Indicates the target sampling point at the time of acquisition. The adjustment factor; This indicates the sampling time corresponding to the cluster center of the target sampling point. The adjustment factor.

[0052] The pressure data at all sampling times within the same cluster determined by the target sampling point are curve-fitted with the flow rate of the corresponding process equipment at all sampling times within the cluster to obtain the numerical relationship between the pressure data and flow rate of the corresponding process equipment, which is denoted as the pressure-flow model. In this model, the pressure data is the independent variable and the flow rate is the dependent variable. The pressure-flow model is combined with the PID parameter tuning method to obtain the baseline value of the PID parameter of the same cluster.

[0053] In this embodiment, polynomial fitting technology is used for curve fitting. Polynomial fitting is a well-known technique and will not be elaborated further. In practical applications, as other implementation methods, while achieving the goal of curve fitting, implementers can use other existing methods such as the least squares method to fit the curve. This application does not impose any special restrictions. In this embodiment, the Ziegler-Nichols method is selected from PID parameter tuning methods to determine the values ​​of the PID parameters. PID parameters include three types: proportional parameter P, integral parameter I, and derivative parameter D. The proportional parameter P is proportional to the error and can quickly reduce the error; the integral parameter I can eliminate steady-state error and stabilize the system; and the derivative parameter D can predict error changes and improve system stability.

[0054] The baseline value of the PID parameter is adjusted based on the difference between the adjustment factors at different acquisition times within the cluster determined by the target sampling point, and the adjusted value of the PID parameter at the acquisition time of the target sampling point is obtained.

[0055] The absolute value of the difference between the adjustment factor of the target sampling point at the acquisition time and the adjustment factor corresponding to the cluster center of the cluster at the acquisition time is denoted as the adjustment ratio of the target sampling point at the acquisition time. The sum of the adjustment ratio of the target sampling point at the acquisition time and the number 1 is denoted as the adjustment magnitude of the target sampling point at the acquisition time. The product of the adjustment magnitude of the target sampling point at the acquisition time and the baseline value of the PID parameter of the cluster at the acquisition time is denoted as the adjustment value of the PID parameter of the target sampling point at the acquisition time.

[0056] It is understandable that for each type of PID parameter, the target sampling point has a corresponding PID parameter adjustment value at the time of acquisition.

[0057] The adjustment value of the PID parameter at the target sampling point at the acquisition time is used as the value of the PID parameter to achieve adaptive control of the flow control valve of the process equipment corresponding to the target sampling point.

[0058] This achieves adaptive control of the flow regulating valve.

[0059] Based on the same inventive concept as the above methods, embodiments of the present invention also provide a PID-based adaptive control system for a flow regulating valve, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described PID-based adaptive control methods for a flow regulating valve.

[0060] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A PID-based adaptive control method for flow regulating valves, characterized in that, The method includes the following steps: Obtain the construction parameters of all semiconductor pipelines, set different sampling points in different pipeline branches, collect pressure data at different sampling times at the sampling points, establish the pressure sequence of the sampling points at the sampling times, and collect the flow rate of the process equipment at the end of each pipeline branch at each sampling time. The last sampling point of any pipeline branch is designated as the target sampling point. Based on the changing trend and relative dispersion of the pressure data of the target sampling point at adjacent sampling times, the first gas pressure characteristic value of the target sampling point at the sampling time is calculated. Any sampling point in all pipeline branches is designated as the sampling point to be analyzed. The energy of the sampling point to be analyzed at the sampling time is calculated. Based on the difference in energy and pressure sequence between the sampling point to be analyzed and its adjacent sampling points at the sampling time, the loss characteristic value of the sampling point to be analyzed at the sampling time is calculated. Based on the loss characteristic values ​​of all sampling points in the pipeline branch where the target sampling point is located at different sampling times, the second gas pressure characteristic value of the target sampling point at the sampling time is calculated. Combining the construction parameters of all pipelines and the first gas pressure characteristic value, the adjustment factor of the target sampling point at the sampling time is calculated. Based on the relative dispersion of the target sampling points at different acquisition times and the adjustment factor, clusters are obtained for all acquisition times of the target sampling points. Based on the pressure data and flow rate of all acquisition times within the same cluster, the baseline value of the PID parameter of the same cluster is determined. The baseline value is adjusted according to the difference between the adjustment factors of different acquisition times within the cluster determined by the target sampling points. The adaptive control of the flow regulating valve is realized based on the adjustment result.

2. The adaptive control method for a flow regulating valve based on PID according to claim 1, characterized in that, The specific method for determining the first gas pressure characteristic value of the target sampling point at the time of sampling is as follows: Any acquisition time is designated as the target acquisition time. Based on the changing trend of pressure data at the target sampling point at the acquisition time adjacent to the target acquisition time, the pressure prediction sequence of the target sampling point at the next adjacent acquisition time is determined. The DTW distance between the pressure prediction sequence of the target sampling point at the next adjacent acquisition time and the pressure prediction sequence of the target acquisition time is designated as the pressure prediction distance of the target sampling point at the target acquisition time. The first preset number of adjacent acquisition times before the target acquisition time are recorded as the previous adjacent times of the target acquisition time. The coefficient of variation of all pressure data of the target sampling point at the target acquisition time and all previous adjacent times of the target acquisition time is recorded as the relative dispersion of the target sampling point at the target acquisition time. The positive correlation between the pressure prediction distance and relative dispersion of the target sampling point at the target acquisition time is recorded as the first gas pressure characteristic value of the target sampling point at the target acquisition time.

3. The adaptive control method for a flow regulating valve based on PID according to claim 2, characterized in that, The specific method for determining the relative dispersion is as follows: The coefficient of variation of all pressure data at the target sampling point at the target acquisition time and at all the time immediately preceding the target acquisition time is denoted as the relative dispersion of the target sampling point at the target acquisition time.

4. The adaptive control method for a flow regulating valve based on PID according to claim 1, characterized in that, The specific method for obtaining the loss characteristic value of the sampling point to be analyzed at the acquisition time is as follows: The sampling point adjacent to the sampling point in the pipeline branch where the sampling point to be analyzed is located is recorded as the previous adjacent sampling point. The absolute value of the energy difference between the sampling point to be analyzed and the previous adjacent sampling point at the time of sampling is recorded as the energy difference of the sampling point to be analyzed at the time of sampling. The ratio of the energy difference of the sampling point to be analyzed at the time of sampling to the energy of the previous adjacent sampling point at the time of sampling is recorded as the energy relative difference of the sampling point to be analyzed at the time of sampling. The DTW distance between the normalized pressure sequence of the sampling point to be analyzed and the previous adjacent sampling point at the acquisition time is denoted as the pressure loss difference of the sampling point to be analyzed at the acquisition time. The positive correlation between the relative energy difference and the pressure loss difference of the sampling point to be analyzed at the time of acquisition is recorded as the loss characteristic value of the sampling point to be analyzed at the time of acquisition.

5. The adaptive control method for a flow regulating valve based on PID according to claim 2, characterized in that, The specific calculation method for the second gas pressure characteristic value of the target sampling point at the time of sampling is as follows: Based on the loss characteristic values ​​of all sampling points in the same pipeline branch at the same acquisition time, establish the column determined by the last sampling point in the same pipeline branch at the same acquisition time. Based on the columns determined by the last sampling point in the same pipeline branch at the target acquisition time and all the previous adjacent times of the target acquisition time, establish the loss characteristic value matrix of the same sampling point at the same acquisition time. Perform singular value decomposition on the loss eigenvalue matrix, and record the maximum value of all singular values ​​as the second gas pressure eigenvalue of the last sampling point in the same pipeline branch at the target sampling time.

6. The adaptive control method for a flow regulating valve based on PID according to claim 1, characterized in that, The formula for calculating the adjustment factor of the target sampling point at the acquisition time is: in, Indicates the first The airflow obstruction of each duct branch; construction parameters include duct length and duct diameter. Indicates the first The first branch of the pipeline The length of each branch pipe; Indicates the first The sum of the lengths of all branch pipes of a given pipeline; Indicates the first The diameter of the main pipeline of each branch pipeline; Indicates the first The first branch of the pipeline The diameter of each branch pipe; Indicates the first The number of all branch pipes included in each pipe branch; Indicates the first The adjustment factor at the time of acquisition for the last sampling point of each pipeline branch; This indicates the preset weighting coefficients; Indicates the first The first gas pressure characteristic value at the time of sampling of the last sampling point of each pipeline branch; Indicates the first The second gas pressure characteristic value at the time of sampling of the last sampling point of each pipeline branch; This represents the tanh function.

7. The adaptive control method for a flow regulating valve based on PID according to claim 3, characterized in that, The specific method for clustering the target sampling points across all sampling times based on the relative dispersion of the target sampling points at different sampling times and the adjustment factor includes: The adjusted metric distance between the target sampling point and the cluster center is used as the clustering metric distance in the FCM algorithm. Clustering is performed on the target sampling point at all sampling times. The formula for calculating the adjusted metric distance between the target sampling point and the cluster center is: in, Indicates the target sampling point at the time of acquisition. Collection time corresponding to cluster center Adjust the distance measurement; Indicates the sampling time corresponding to the cluster center of the target sampling point. The relative dispersion of the pressure sequence; Indicates the target sampling point at the time of acquisition. The relative dispersion of the pressure sequence; Indicates the target sampling point at the time of acquisition. The adjustment factor; Indicates the sampling time corresponding to the cluster center of the target sampling point. The adjustment factor.

8. The adaptive control method for a flow regulating valve based on PID according to claim 1, characterized in that, The specific method for obtaining the baseline values ​​of the PID parameters for the same cluster is as follows: The pressure data at all sampling times within the same cluster determined by the target sampling point are fitted with the flow curves of the process equipment corresponding to the target sampling point at all sampling times within the cluster to obtain the numerical relationship between the pressure data and flow of the process equipment corresponding to the target sampling point. Combined with the PID parameter tuning method, the baseline value of the PID parameter of the same cluster is obtained.

9. The adaptive control method for a flow regulating valve based on PID according to claim 1, characterized in that, The method for adjusting the baseline value based on the difference in adjustment factors between different sampling times within the cluster determined by the target sampling point, and achieving adaptive control of the flow regulating valve based on the adjustment result, includes the following specific methods: The sum of the absolute value of the difference between the adjustment factor of the target sampling point at the acquisition time and the adjustment factor at the acquisition time corresponding to the cluster center of the cluster where the acquisition time is located, and the number 1, is recorded as the adjustment range of the target sampling point at the acquisition time. The product of the adjustment range of the target sampling point at the acquisition time and the baseline value of the PID parameter of the cluster where the acquisition time is located is recorded as the adjustment value of the PID parameter of the target sampling point at the acquisition time. The adjustment value of the PID parameter at the target sampling point at the acquisition time is used as the value of the PID parameter to achieve adaptive control of the flow control valve of the process equipment corresponding to the target sampling point.

10. A PID-based adaptive control system for a flow regulating valve, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as claimed in any one of claims 1-9.

Images