Equipment cleaning effect evaluation method based on real-time data simulation model

Abstract

The present invention discloses an equipment cleaning effect evaluation method based on a real-time data simulation model, and relates to the technical field of semiconductor equipment cleaning. By detecting the medium concentration of a cleaning medium used to clean equipment parts in a cleaning project, it is judged whether the cleaning effect of the equipment parts is sufficient when the cleaning project is completed according to the change of the medium concentration. When the cleaning effect of the equipment parts is not sufficient, the medium concentration required for secondary cleaning is predicted according to the change of the medium concentration, and a secondary cleaning strategy is obtained according to the prediction result, thereby realizing full-automatic monitoring and regulation of the equipment parts during the cleaning process, thereby reducing the occurrence of insufficient cleaning of equipment parts caused by human operation factors, and solving the problems of low process efficiency and insufficient cleaning of equipment parts caused by frequent human operation errors.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor equipment cleaning, and in particular to an equipment cleaning effect evaluation method based on a real-time data simulation model. Background Art

[0002] Semiconductor equipment operates in this environment for a long time, and the core parts inside the equipment are prone to adhere to many corrosive pollutants or particles. In order to improve the stability and service life of semiconductor equipment, its parts need to be cleaned;

[0003] In the prior art, for the cleaning of equipment parts, a cleaning medium of standard concentration is often used to clean the equipment parts for a fixed time. This method still requires manual inspection of the cleaning effect of the equipment parts after cleaning, and then determines whether a secondary cleaning is needed based on experience. This process is inefficient and often results in insufficient cleaning of equipment parts due to human operating errors. How to effectively evaluate the cleaning effect of equipment parts while providing a reasonable secondary cleaning strategy when the equipment parts need secondary cleaning is a problem we need to solve. To this end, a method for evaluating the cleaning effect of equipment parts based on a real-time data simulation model is provided. Summary of the invention

[0004] The purpose of the present invention is to provide an equipment cleaning effect evaluation method based on a real-time data simulation model.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] The equipment cleaning effect evaluation method based on the real-time data simulation model includes the following steps:

[0007] Step S1: constructing a device cleaning sample database, and constructing a simulation environment space linked to the device cleaning sample database;

[0008] Step S2: acquiring equipment cleaning data in real time, and importing the acquired equipment cleaning data into the simulation environment space;

[0009] Step S3: constructing a corresponding target part simulation model in the simulation environment space according to the imported equipment cleaning data;

[0010] Step S4: Evaluate the cleaning effect of the constructed target part simulation model, and formulate a secondary cleaning strategy based on the evaluation results.

[0011] Furthermore, the process of constructing the device cleaning sample database includes:

[0012] Set up a number of standard target parts, and build corresponding data storage space in the equipment cleaning sample database according to each standard target part;

[0013] The standard target part is composed of a number of standard parts, and each standard part is associated with at least one cleaning item;

[0014] Setting corresponding standard cleaning data for each cleaning item, wherein the standard cleaning data includes a cleaning medium and a cleaning method;

[0015] Importing into each data storage space the size information and contour information of the corresponding standard target part and the standard parts constituting the standard target part;

[0016] Construct a standard target part model and a standard component model corresponding to the standard target part according to the size information and the contour information;

[0017] Import the constructed standard target part model and standard component model into the corresponding data storage space.

[0018] Furthermore, the construction process of the simulation environment space construction includes:

[0019] Constructing corresponding data transmission links according to the data storage spaces, and linking each data storage space with the simulation environment space through the constructed data transmission links;

[0020] A data loading link is constructed to communicate with a data acquisition terminal for acquiring equipment cleaning data, and the equipment cleaning data acquired in real time by the data acquisition terminal is imported into the simulation environment space through the data loading link.

[0021] Furthermore, the process of acquiring the equipment cleaning data includes:

[0022] Record the equipment that needs to be cleaned as the target part;

[0023] Obtain all cleaning processes for cleaning the target part and generate corresponding cleaning process nodes;

[0024] Set several cleaning items for each cleaning process node, and set corresponding data collection terminals according to the set cleaning items;

[0025] Acquiring equipment cleaning data from the start time to the end time of the cleaning project through the set data acquisition terminal; the equipment cleaning data includes the name of equipment parts, cleaning medium and medium concentration;

[0026] The equipment cleaning data from the start time to the end time of each cleaning project is packaged, and the packaged data is imported into the simulation environment space.

[0027] Furthermore, the process of constructing the target part simulation model includes:

[0028] Construct a three-dimensional coordinate system in the simulation environment space, and retrieve the corresponding standard target part model in the equipment cleaning sample database according to the collected equipment cleaning data;

[0029] Mapping the retrieved standard target part model into a three-dimensional coordinate system; the standard target part model includes a plurality of equipment parts, and different equipment parts are associated with at least one cleaning item;

[0030] According to the current cleaning project, the corresponding equipment parts associated with the cleaning project in the standard target part model are highlighted;

[0031] and indexing the standard component model corresponding to the highlighted equipment component;

[0032] Constructing a two-dimensional coordinate system of time with respect to medium concentration associated with a standard component model;

[0033] Generate a corresponding concentration change curve according to the medium concentration in the equipment cleaning data from the start time to the end time of the cleaning project;

[0034] Construct a two-dimensional coordinate system of time and concentration for the highlighted equipment parts;

[0035] The generated concentration change curve is mapped into a two-dimensional coordinate system to obtain a medium state change diagram.

[0036] Furthermore, the process of evaluating the cleaning effect of the constructed target part simulation model includes:

[0037] Setting a plurality of sampling points on the concentration change curve in the obtained medium state change diagram;

[0038] Obtaining the medium concentration value at the sampling point on the concentration variation curve;

[0039] According to the medium concentration value of each sampling point, the node concentration change rate of the corresponding sampling point is obtained;

[0040] Constructing a node sequence table, and setting a sequence item corresponding to the sampling point in the constructed node sequence table;

[0041] Import the obtained node concentration change rate of each sampling point into the corresponding sequence item;

[0042] Generate an index pointer, and set a corresponding pointer vector according to the relative relationship between each sequence item of the index pointer;

[0043] The relative relationship includes a progressive relationship and a balanced relationship;

[0044] The pointer vector includes a progressive pointer vector and a balanced pointer vector;

[0045] The number of balance pointer vectors in the node sequence table is obtained. If the number of balance pointer vectors is 0, it indicates that the cleaning of the equipment parts is insufficient, and a secondary cleaning requirement instruction is generated;

[0046] If the number of the balance pointer vector is not 0, the sampling point corresponding to the sequence item where the balance pointer vector is located is marked. If the time corresponding to the sampling point is not the end time of the cleaning project, it is obtained whether the medium concentration value corresponding to the sampling point is 0. If it is not 0, it means that the equipment parts cleaned by the cleaning project are sufficiently cleaned. If it is 0, it means that the equipment parts are not sufficiently cleaned, and a secondary cleaning demand instruction is generated;

[0047] A secondary cleaning strategy is formulated based on the generated secondary cleaning demand instructions for the equipment parts cleaned by the cleaning project.

[0048] Further, the process of formulating a secondary cleaning strategy for the equipment parts cleaned by the cleaning project according to the generated secondary cleaning requirement instruction includes:

[0049] Mark each sampling point corresponding to the obtained sequence item with a progressive relationship, and obtain the corresponding node concentration change rate;

[0050] Get the vector value corresponding to each pointer vector;

[0051] Generate a corresponding vector change curve according to the obtained vector value, and generate a prediction curve part according to the generated vector change curve, wherein the prediction curve part includes at least one prediction sampling point, and the interval length of the prediction sampling point is t;

[0052] According to the number of predicted sampling points generated, the corresponding predicted cleaning time is obtained, and according to the predicted cleaning time, the medium concentration and cleaning time required for the corresponding secondary cleaning are obtained.

[0053] Furthermore, the generation process of the prediction curve part includes:

[0054] Label each vector value, denoted as j, where j = 1, 2, ... m, m is an integer, and m> 1;

[0055] The vector values ​​labeled j=m and j=m-1 are respectively denoted as Ex m and Ex m-1 ;

[0056] The generated prediction sampling point is recorded as k, where k = 1, 2, ..., s, s is an integer, and s ≥ 1;

[0057] Then the prediction vector value corresponding to the prediction sampling point corresponding to label k=1 is recorded as Yx1;

[0058] Then Yx1=2Ex m-1 -Ex m ;

[0059] If Yx1≤0, no new prediction sampling point is generated;

[0060] If Yx1>0, generate a new prediction sampling point and record it as Yx2;

[0061] Then Yx2=2Yx1-Ex m-1 ;

[0062] If Yx2≤0, no new prediction sampling point is generated;

[0063] If Yx2>0, generate a new prediction sampling point and record it as Yx3, and so on;

[0064] Connect all prediction sampling points in sequence to obtain the corresponding prediction curve part.

[0065] Compared with the prior art, the beneficial effects of the present invention are as follows: by detecting the medium concentration of the cleaning medium used to clean equipment parts in the cleaning project, it is judged whether the cleaning effect of the equipment parts is sufficient when the cleaning project is completed according to the change of the medium concentration; and when the cleaning effect of the equipment parts is not sufficient, the medium concentration required for secondary cleaning is predicted according to the change of the medium concentration, and the secondary cleaning strategy is obtained according to the prediction result, thereby realizing fully automatic monitoring and regulation of the cleaning process of equipment parts, thereby reducing the occurrence of insufficient cleaning of equipment parts caused by human operation factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0067] like Figure 1 As shown, the equipment cleaning effect evaluation method based on the real-time data simulation model includes the following steps:

[0068] Step S1: constructing a device cleaning sample database, and constructing a simulation environment space linked to the device cleaning sample database;

[0069] Step S2: acquiring equipment cleaning data in real time, and importing the acquired equipment cleaning data into the simulation environment space;

[0070] Step S3: constructing a corresponding target part simulation model in the simulation environment space according to the imported equipment cleaning data;

[0071] Step S4: Evaluate the cleaning effect of the constructed target part simulation model, and formulate a secondary cleaning strategy based on the evaluation results.

[0072] It should be further explained that, in the specific implementation process, the process of constructing the device cleaning sample database includes:

[0073] Set up a number of standard target parts, and build corresponding data storage space in the equipment cleaning sample database according to each standard target part;

[0074] It should be further explained that, in the specific implementation process, the data storage space is only associated with a single standard target part;

[0075] The standard target part is composed of a number of standard parts, and each standard part is associated with at least one cleaning item;

[0076] Corresponding standard cleaning data is set for each cleaning item, and the standard cleaning data includes a cleaning medium and a cleaning method; it should be further explained that, in the specific implementation process, the cleaning medium includes a liquid medium and a gas medium, and the cleaning method includes soaking, rinsing, and blowing dry;

[0077] Importing into each data storage space the size information and contour information of the corresponding standard target part and the standard parts constituting the standard target part;

[0078] Construct a standard target part model and a standard component model corresponding to the standard target part according to the size information and the contour information;

[0079] Import the constructed standard target part model and standard component model into the corresponding data storage space.

[0080] It should be further explained that, in the specific implementation process, the construction process of the simulation environment space construction includes:

[0081] Constructing corresponding data transmission links according to the data storage spaces, and linking each data storage space with the simulation environment space through the constructed data transmission links;

[0082] A data loading link is constructed to communicate with a data acquisition terminal for acquiring equipment cleaning data, and the equipment cleaning data acquired in real time by the data acquisition terminal is imported into the simulation environment space through the data loading link.

[0083] It should be further explained that, in the specific implementation process, the process of obtaining the equipment cleaning data includes:

[0084] Record the equipment that needs to be cleaned as the target part;

[0085] Obtain all cleaning processes for cleaning the target part and generate corresponding cleaning process nodes;

[0086] Set several cleaning items for each cleaning process node, and set corresponding data collection terminals according to the set cleaning items;

[0087] Acquiring equipment cleaning data from the start time to the end time of the cleaning project through the set data acquisition terminal; the equipment cleaning data includes the name of equipment parts, cleaning medium and medium concentration;

[0088] The equipment cleaning data from the start time to the end time of each cleaning project is packaged, and the packaged data is imported into the simulation environment space.

[0089] It should be further explained that, in the specific implementation process, the process of constructing the target part simulation model includes:

[0090] Construct a three-dimensional coordinate system in the simulation environment space, and retrieve the corresponding standard target part model in the equipment cleaning sample database according to the collected equipment cleaning data;

[0091] Mapping the retrieved standard target part model into a three-dimensional coordinate system; it should be further explained that, in a specific implementation process, the standard target part model includes a plurality of equipment parts, and different equipment parts are associated with at least one cleaning item;

[0092] According to the current cleaning project, the corresponding equipment parts associated with the cleaning project in the standard target part model are highlighted;

[0093] and indexing the standard component model corresponding to the highlighted equipment component;

[0094] Constructing a two-dimensional coordinate system of time with respect to medium concentration associated with a standard component model;

[0095] Generate a corresponding concentration change curve according to the medium concentration in the equipment cleaning data from the start time to the end time of the cleaning project;

[0096] Construct a two-dimensional coordinate system of time and concentration for the highlighted equipment parts;

[0097] The generated concentration change curve is mapped into a two-dimensional coordinate system to obtain a medium state change diagram.

[0098] It should be further explained that, in the specific implementation process, the process of evaluating the cleaning effect of the constructed target part simulation model includes:

[0099] A plurality of sampling points are set on the concentration change curve in the obtained medium state change diagram, and each sampling point is marked as i, wherein i=1, 2, ..., n, n is an integer, and n>2;

[0100] In the specific implementation process, the time corresponding to the sampling point with label i=1 is the start time of the cleaning project, and the time corresponding to the sampling point with label i=n is the end time of the cleaning project, and the interval between each adjacent sampling point is the same, which is t;

[0101] Obtain the medium concentration value at the sampling point labeled i on the concentration change curve, and mark the medium concentration value as Jn i ;

[0102] According to the medium concentration value of each sampling point, the node concentration change rate of the corresponding sampling point is obtained, which is recorded as Jb i+1 ,in

[0103] Jb i+1 =(Jn i -Jn i+1 ) / t;

[0104] It should be further explained that, in the specific implementation process, the node concentration change rate is not calculated for the sampling point labeled i = 1;

[0105] Constructing a node sequence table, and setting sequence items corresponding to the sampling points in the constructed node sequence table; it should be further explained that, in a specific implementation process, each of the sequence items is associated with a sampling point, and the arrangement order of the sequence items is the same as the label order of the sampling points;

[0106] Import the obtained node concentration change rate of each sampling point into the corresponding sequence item;

[0107] Generate an index pointer, and set a corresponding pointer vector according to the relative relationship between each sequence item of the index pointer;

[0108] The relative relationship includes a progressive relationship and a balanced relationship;

[0109] The pointer vector includes a progressive pointer vector and a balanced pointer vector;

[0110] When the node concentration change rate in the sequence item associated with the sampling point labeled i is greater than the node concentration change rate in the sequence item associated with the sampling point labeled i+1, the relative relationship between the two sequence items is a progressive relationship, and a corresponding progressive pointer vector is generated;

[0111] When the node concentration change rate in the sequence item associated with the sampling point labeled i is equal to the node concentration change rate in the sequence item associated with the sampling point labeled i+1, the relative relationship between the two sequence items is a balanced relationship, and a corresponding balanced pointer vector is generated;

[0112] The number of balance pointer vectors in the node sequence table is obtained. If the number of balance pointer vectors is 0, it indicates that the cleaning of the equipment parts is insufficient, and a secondary cleaning requirement instruction is generated;

[0113] If the number of the balance pointer vector is not 0, the sampling point corresponding to the sequence item where the balance pointer vector is located is marked. If the time corresponding to the sampling point is not the end time of the cleaning project, it is obtained whether the medium concentration value corresponding to the sampling point is 0. If it is not 0, it means that the equipment parts cleaned by the cleaning project are sufficiently cleaned. If it is 0, it means that the equipment parts are not sufficiently cleaned, and a secondary cleaning demand instruction is generated;

[0114] A secondary cleaning strategy is formulated based on the generated secondary cleaning demand instructions for the equipment parts cleaned by the cleaning project.

[0115] It should be further explained that, in the specific implementation process, the process of formulating a secondary cleaning strategy for the equipment parts cleaned by the cleaning project according to the generated secondary cleaning demand instruction includes:

[0116] Mark each sampling point corresponding to the obtained sequence item with a progressive relationship, and obtain the corresponding node concentration change rate;

[0117] Obtain the vector value corresponding to each pointer vector; it should be further explained that, in the specific implementation process, the vector value of the progressive pointer vector is the difference in the node concentration change rate of the sampling points corresponding to the sequence items with a progressive relationship;

[0118] Generate a corresponding vector change curve according to the obtained vector value, and generate a prediction curve part according to the generated vector change curve, wherein the prediction curve part includes at least one prediction sampling point; it should be further explained that, in a specific implementation process, when the vector value corresponding to the prediction sampling point is less than or equal to 0, the prediction sampling point is not generated further; it should be further explained that, in a specific implementation process, the interval time length of the prediction sampling points is t;

[0119] According to the number of predicted sampling points generated, the corresponding predicted cleaning time is obtained, and according to the predicted cleaning time, the medium concentration and cleaning time required for the corresponding secondary cleaning are obtained;

[0120] It should be further explained that, in the specific implementation process, the generation process of the prediction curve part includes:

[0121] Label each vector value, denoted as j, where j = 1, 2, ... m, m is an integer, and m> 1;

[0122] The vector values ​​labeled j=m and j=m-1 are respectively denoted as Ex m and Ex m-1 ;

[0123] The generated prediction sampling point is recorded as k, where k = 1, 2, ..., s, s is an integer, and s ≥ 1;

[0124] Then the prediction vector value corresponding to the prediction sampling point corresponding to label k=1 is recorded as Yx1;

[0125] Then Yx1=2Ex m-1 -Ex m ;

[0126] If Yx1≤0, no new prediction sampling point is generated;

[0127] If Yx1>0, generate a new prediction sampling point and record it as Yx2;

[0128] Then Yx2=2Yx1-Ex m-1 ;

[0129] If Yx2≤0, no new prediction sampling point is generated;

[0130] If Yx2>0, generate a new prediction sampling point and record it as Yx3, and so on;

[0131] Connect all prediction sampling points in sequence to obtain the corresponding prediction curve part.

[0132] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. The equipment cleaning effect evaluation method based on the real-time data simulation model is characterized by: The following steps are involved: Step S1: constructing a device cleaning sample database, and constructing a simulation environment space linked to the device cleaning sample database; Step S2: acquiring equipment cleaning data in real time, and importing the acquired equipment cleaning data into the simulation environment space; Step S3: constructing a corresponding target part simulation model in the simulation environment space according to the imported equipment cleaning data; Step S4: Evaluate the cleaning effect of the constructed target part simulation model, and formulate a secondary cleaning strategy based on the evaluation results; The process of constructing the equipment cleaning sample database includes: Set up a number of standard target parts, and build corresponding data storage space in the equipment cleaning sample database according to each standard target part; The standard target part is composed of a number of standard parts, and each standard part is associated with at least one cleaning item; Setting corresponding standard cleaning data for each cleaning item, wherein the standard cleaning data includes a cleaning medium and a cleaning method; Importing into each data storage space the size information and contour information of the corresponding standard target part and the standard parts constituting the standard target part; Construct a standard target part model and a standard component model corresponding to the standard target part according to the size information and the contour information; Import the constructed standard target part model and standard component model into the corresponding data storage space; The construction process of the simulation environment space construction includes: Constructing corresponding data transmission links according to the data storage spaces, and linking each data storage space with the simulation environment space through the constructed data transmission links; Constructing a data loading link that is in communication with a data acquisition terminal for acquiring equipment cleaning data, and importing the equipment cleaning data acquired in real time by the data acquisition terminal into the simulation environment space through the data loading link; The process of obtaining the equipment cleaning data includes: Record the equipment that needs to be cleaned as the target part; Obtain all cleaning processes for cleaning the target part and generate corresponding cleaning process nodes; Set several cleaning items for each cleaning process node, and set corresponding data collection terminals according to the set cleaning items; Acquiring equipment cleaning data from the start time to the end time of the cleaning project through the set data acquisition terminal; the equipment cleaning data includes the name of equipment parts, cleaning medium and medium concentration; Packaging the equipment cleaning data from the start time to the end time of each cleaning project, and importing the packaged data into the simulation environment space; The process of constructing the target part simulation model includes: Construct a three-dimensional coordinate system in the simulation environment space, and retrieve the corresponding standard target part model in the equipment cleaning sample database according to the collected equipment cleaning data; Mapping the retrieved standard target part model into a three-dimensional coordinate system; the standard target part model includes a plurality of equipment parts, and different equipment parts are associated with at least one cleaning item; According to the current cleaning project, the corresponding equipment parts associated with the cleaning project in the standard target part model are highlighted; and indexing the standard component model corresponding to the highlighted equipment component; Constructing a two-dimensional coordinate system of time with respect to medium concentration associated with a standard component model; Generate a corresponding concentration change curve according to the medium concentration in the equipment cleaning data from the start time to the end time of the cleaning project; Construct a two-dimensional coordinate system of time and concentration for the highlighted equipment parts; Mapping the generated concentration change curve into a two-dimensional coordinate system to obtain a medium state change diagram; The process of evaluating the cleaning effect of the constructed target part simulation model includes: Setting a plurality of sampling points on the concentration change curve in the obtained medium state change diagram; Obtaining the medium concentration value at the sampling point on the concentration variation curve; According to the medium concentration value of each sampling point, the node concentration change rate of the corresponding sampling point is obtained; Constructing a node sequence table, and setting a sequence item corresponding to the sampling point in the constructed node sequence table; Import the obtained node concentration change rate of each sampling point into the corresponding sequence item; Generate an index pointer, and set a corresponding pointer vector according to the relative relationship between each sequence item of the index pointer; The relative relationship includes a progressive relationship and a balanced relationship; The pointer vector includes a progressive pointer vector and a balanced pointer vector; The number of balance pointer vectors in the node sequence table is obtained. If the number of balance pointer vectors is 0, it indicates that the cleaning of the equipment parts is insufficient, and a secondary cleaning requirement instruction is generated; If the number of the balance pointer vector is not 0, the sampling point corresponding to the sequence item where the balance pointer vector is located is marked. If the time corresponding to the sampling point is not the end time of the cleaning project, it is obtained whether the medium concentration value corresponding to the sampling point is 0. If it is not 0, it means that the equipment parts cleaned by the cleaning project are sufficiently cleaned. If it is 0, it means that the equipment parts are not sufficiently cleaned, and a secondary cleaning demand instruction is generated; Formulate a secondary cleaning strategy for the equipment parts cleaned by the cleaning project according to the generated secondary cleaning demand instructions; The process of formulating a secondary cleaning strategy for the equipment parts cleaned by the cleaning project according to the generated secondary cleaning demand instructions includes: Mark each sampling point corresponding to the obtained sequence item with a progressive relationship, and obtain the corresponding node concentration change rate; Get the vector value corresponding to each progressive pointer vector; Generate a corresponding vector change curve according to the obtained vector value, and generate a prediction curve part according to the generated vector change curve, wherein the prediction curve part includes at least one prediction sampling point, and the interval length of the prediction sampling point is t; According to the number of predicted sampling points generated, the corresponding predicted cleaning time is obtained, and according to the predicted cleaning time, the medium concentration and cleaning time required for the corresponding secondary cleaning are obtained.

2. The equipment cleaning effect evaluation method based on the real-time data simulation model according to claim 1 is characterized in that: The generation process of the prediction curve part includes: Label each vector value, denoted as j, where j=1, 2, ...m, m is an integer, and m>1; The vector values ​​labeled j=m and j=m-1 are respectively denoted as Ex m and Ex m-1 ; The generated prediction sampling point is recorded as k, where k=1, 2, ..., s, s is an integer, and s≥1; Then the prediction vector value corresponding to the prediction sampling point corresponding to label k=1 is recorded as Yx1; Then Yx1=2Ex m-1 -Ex m ; If Yx1≤0, no new prediction sampling point is generated; If Yx1>0, generate a new prediction sampling point and record it as Yx2; Then Yx2=2Yx1-Ex m-1 ; If Yx2≤0, no new prediction sampling point is generated; If Yx2>0, generate a new prediction sampling point and record it as Yx3, and so on; Connect all prediction sampling points in sequence to obtain the corresponding prediction curve part.

Images