WO2023226216A1

WO2023226216A1 - Smart rank downlink rate optimization method applicable to 6g

Info

Publication number: WO2023226216A1
Application number: PCT/CN2022/114212
Authority: WO
Inventors: 王玉梁
Original assignee: 中电信数智科技有限公司
Priority date: 2022-05-25
Filing date: 2022-08-23
Publication date: 2023-11-30
Also published as: CN115052296A; CN115052296B

Abstract

Disclosed in the present invention is a smart Rank downlink rate optimization method applicable to 6G, comprising: classifying affecting factors for a Rank, and on the basis of a regional prediction model, obtaining a drive test Rank downlink rate fault occurrence probability prediction value and obtaining classification identifiers for factors affecting the Rank; on the basis of a random forest model, analyzing data of four scenes covered by antennas and factor classification identifiers in Rank identifiers so as to obtain the biggest factor for a Rank fault which most likely occurs in a scene/sub-scene, positioning a region where a UE is located, generating a region identifier, and generating a biggest factor identifier for the Rank fault of the scene/sub-scene; and constructing a Rank optimization program and associating same with the scenes covered by the antennas, so that self-healing adjustment and optimization can be realized after a fault occurs in a 5G Rank link during a drive test downlink rate optimization process of a UE. The present invention can support 5G wireless downlink rate optimization and emerging 6G wireless downlink rate optimization during the process of transition from 5G to 6G.

Description

A method for intelligent Rank downlink rate optimization that can be used for 6G

Technical field

The invention belongs to the field of wireless communication technology, and specifically relates to a method that can be used for 6G intelligent Rank downlink rate optimization.

Background technique

With the gradual advancement of 5G network construction, the gap between the current 5G network scale and 4G is getting smaller and smaller. More 4G users are migrating to 5G networks. How 5G networks can provide users with a better experience than 4G networks has become a challenge currently faced by operators. main problem. 5G network is an extension of 4G network. Simply good coverage can no longer reflect the value advantage of 5G network. The advantage of 5G terminals over 4G terminals is that they have more antenna port support (the mainstream configuration of 5G commercial is 2T4R).

In the communication system, the UE (terminal) performs channel estimation based on the wireless CSI-RS reference signal and calculates the maximum number of flows with the smallest downlink channel coherence, which is called "Rank", which is translated as "Rank" in Chinese. The UE uses CSI to RI (Rank indicator) is reported to the base station. The 5G Rank (stream) has a great impact on the user rate. Under good coverage, a low Rank will directly lead to a low rate. The user perception is the same as that of 4G. If the Rank has one more stream, the downlink rate can be more than 100 Mbit/s.

Taking advantage of the feature of 5G terminals that can support more streams, increasing the number of 5G user Rank streams has become the key to 5G network optimization. The quality of the network experience ultimately depends on the Rank gap. Since Rank is greatly affected by factors such as the wireless environment, terminal antennas, AAU antennas, and channel correlations, and single-point analysis during the Rank optimization process is time-consuming and labor-intensive, how to quickly and accurately improve 5G Rank has become the main research object of current and future optimization. .

Contents of the invention

The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned existing technologies and provide a method for optimizing the intelligent Rank downlink rate for 6G, which can support the 5G wireless downlink to the emerging 6G wireless downlink rate during the transition from 5G to 6G. optimization.

In order to achieve the above technical objectives, the technical solutions adopted by the present invention are:

A method for intelligent Rank downlink rate optimization that can be used for 6G, which is characterized by including:

Step 1: Combine the log data to classify the influencing factors of Rank. Based on the regional prediction model, obtain the predicted value of the failure probability of the drive test Rank downlink rate, obtain the classification identifier of the factors that affect the Rank, and store the factor classification identifier and the predicted value of the failure probability. Rank identification;

Step 2: Based on the random forest model, analyze the four scenario data of antenna coverage stored in the gNB log and the factor classification identifier in the Rank identifier, to obtain the most likely Rank failure factors in the scenario/sub-scenario, and at the same time analyze the UE Locate the area and generate the area identifier, and finally generate the factor identifier with the largest scene/sub-scene Rank failure, and store the area identifier, scene identifier and factor identifier into the Rank identifier;

Step 3: Based on the 5G Rank problem optimization measures and Rank identification analysis, build a Rank optimization program and associate it with the antenna coverage scenario/sub-scenario to realize the self-healing adjustment and adjustment after the 5G Rank link fails during the UE's drive test downlink rate optimization process. optimization.

In order to optimize the above technical solutions, specific measures taken also include:

The above steps analyze the historical log data stored on the base station equipment to find out the data of factors that affect Rank and classify them according to hardware, UE placement, base station RF, and algorithm;

The most important factor data in each category is input into the constructed regional prediction model for analysis, and the sub-category failure probability prediction indicators of each category of factors that affect Rank are obtained, and the corresponding factor classification identification is generated.

The regional prediction model described in step 1 above is:

P(A|B)=(P(B|A)*P(A))/P(B|A)P(A)+P(B|A')P(A')

P(A|B) is the probability of abnormality in factors affecting Rank;

P(B|A) is the probability of the result of the total number of abnormal items of factors affecting Rank/the total number of historical database items during the continuous learning process of the regional prediction model;

P(A) is the total number of data anomalies/the total number of historical data that ignores other factors and factors that affect Rank;

P(B|A') is the probability that the total number of data anomalies that affect Rank has appeared in the historical database;

P(A')=1-P(A).

In the above step one, the sub-category failure probability prediction indicators and corresponding factor classification identifiers of each category are as follows:

The specific classification indicators of its prediction indicators and factors are as follows:

Factor 1: Hardware, corresponding factor classification identification: 2-1;

Prediction index: whether the channel correction is passed, corresponding factor classification identification: 2-1-1;

Factor 2: UE placement, corresponding factor classification identification: 2-2;

Prediction indicator 1: RSRP balance between UE antennas, corresponding factor classification identification 1: 2-2-1;

Prediction indicator 2: UE placement position and method, corresponding factor classification identification 2: 2-2-2;

Factor three: base station RF, corresponding factor classification identification: 2-3;

Predictive indicator 1: Toward the building, the direction angle of the reflection path is increased, corresponding to factor classification identification 1: 2-3-1;

Prediction indicator 2: Open scenes increase the downtilt angle of ground reflection, corresponding to factor classification identification 2: 2-3-2;

Factor 4: Algorithm, corresponding factor classification identifier: 2-4;

Prediction indicator 1: Tianxuan terminal: SRS right, corresponding factor classification identification 1: 2-4-1;

Prediction indicator 2: Unselected terminal: VAM+PMI right, corresponding factor classification identification 2: 2-4-2.

In the above step 1, the factor classification identifier and the predicted value of failure probability are combined into the Rank identifier using the # symbol. The format is: factor classification identifier #predicted value of probability of failure.

The above step 2 builds the random forest model as follows:

Among them, n is the number of sampling scenes, which is 4, indicating that the sampling scenes include scenes one, two, three, and four;

|Di|/|D| refers to the probability of scenarios one, two, three and four;

H(i) is the total number of features in scenes one, two, three and four.

The above scenes one, two, three and four and their scene identifiers are:

Scenario 1: The wireless environment is single and there are few buildings. The main scene identification is: 1-1;

The second scene: the road is narrow, with buildings on both sides in groups, the main scene logo: 1-2;

Scene 3: Multi-lane intersection, clusters of buildings, open road space, main scene identification: 1-3;

Scene 4: Multi-lane road, single rows of buildings, shady trees, scene identification: 1-4.

In the above step 2, combine the area identifier, scene identifier, and factor identifier with the # symbol and put them into the Rank identifier. The format is: area identifier # scene identifier # factor identifier # factor classification identifier # failure probability prediction value.

The above step three is optimizing the downlink rate during the drive test. When an alarm occurs in the Rank link, the processing details are as follows:

S1, split the Rank identifier to obtain the area where the UE is located;

S2, split the Rank identifier to obtain the corresponding scene/sub-scene;

S3: Extract the fault keyword based on the abnormal content of the drive test data report and compare it with the factor identifier obtained by splitting the Rank identifier. If the fault keyword appears in the abnormal content one or more times, it is confirmed that the abnormality is real and valid;

S4: Split the Rank identifier to obtain the predicted value of the probability of failure. If the predicted value of the probability is greater than 50%, execute the corresponding preset Rank optimization program to achieve self-healing or optimization of the fault.

The invention has the following beneficial effects:

In the process of 5G downlink rate optimization, the idea of Rank optimization is to increase multipath and reduce spatial correlation. Optimizing the wireless environment is a key means to improve Rank. This invention combines 5G base station scheduling parameters and 6G new feature parameters to maximize Rank while reducing inter-beam interference and increasing the overall system capacity.

In 6G space scenarios, due to the widespread application of terahertz and millimeter waves, the coverage environment has been greatly improved compared to 5G. The present invention further uses an artificial intelligence model combined with the road test downlink rate optimization process of 5G under the premise that there is currently no practical application of 6G. When the Rank link fails, the efficient application of artificial intelligence models makes up for the efficiency of 6G optimization and adjustment in low-altitude areas, terahertz, millimeter waves and scenarios. This solves the difficulty of optimizing 5G Rank for the spatial coverage environment.

Description of the drawings

Figure 1 shows the sub-classification failure probability prediction indicators of each classification in the present invention;

Figure 2 shows the design principle of the random forest model of the present invention;

Figure 3 is a schematic diagram of the intelligent Rank downlink rate optimization method applicable to 6G according to the present invention.

Detailed ways

The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

As shown in Figure 3, a method of the present invention that can be used for 6G intelligent Rank downlink rate optimization mainly includes three parts:

Step 1: Combine the log data to classify the influencing factors of Rank. Based on the regional prediction model, obtain the predicted value of the failure probability of the road test Rank downlink rate, obtain the classification identifier of the factors that affect the Rank, and store the factor classification identifier and the predicted value of the failure probability. Rank identification;

That is to build a regional prediction model, and combine the influencing factors of the Rank of the UE regional positioning prediction area with the downlink rate failure probability of each classification drive test: Construct a [regional prediction model], and the difficulty of 5G Rank optimization is to optimize the spatial coverage environment. When the UE When optimizing the downlink rate of the drive test and encountering a low Rank situation, the constructed [regional prediction model] is combined with the log data to classify the factors affecting the Rank, and then the classification of factors affecting the Rank and the probability of failure probability of the downlink rate of the drive test Rank are obtained. value, and generate factor classification identifiers at the same time. And store the factor classification identification and failure probability prediction value in the Rank identification using a combination of # symbols. The format is: factor classification identification # failure probability prediction value.

specific description:

Step 1: Analyze the historical log data stored on the base station equipment to find out the data of factors that affect Rank and classify them according to hardware, UE placement, base station RF, and algorithm;

The most important factor data in each classification is input into the constructed regional prediction model for analysis, and the sub-classification failure probability prediction indicators of each classification of factors that affect Rank are obtained, and the corresponding factor classification identification is generated, as shown in Figure 1.

Build [Regional Forecasting Model]

Formula: P(A|B)=(P(B|A)*P(A))/P(B|A)P(A)+P(B|A')P(A')

[Prior probability] = P (A) [Conditional probability] = P (B) [Adjustment factor] = P (B | A) x P (A)

P(A) is the total number of data anomalies/total number of historical data that affects Rank regardless of other factors, for example: 40%;

P(A')1-P(A), here it is 60%;

P(B|A) [Regional Prediction Model] In the continuous learning process, the probability of the total number of abnormal items of factors affecting Rank/the total number of historical database items is 50% here;

P(B|A') The probability that the total number of data anomalies that affect Rank has appeared in the historical database. If the historical database has normal energy consumption, the default is 100%;

P(B) is an abnormality probability formula that ignores other factors and directly considers the factors that affect Rank;

P(B)=P(B|A)P(A)+P(B|A')P(A'), here it is 0.5*0.4+1*0.6=0.8;

Then according to Bayes' formula, it can be calculated, that is:

P(A|B)＝(0.5*0.4)/(0.8)＝0.25

0.25 is the probability of abnormality in the factors that affect Rank, thus completing the calculation of the entire [Regional Prediction Model]. The sub-category failure probability prediction indicators and corresponding factor classification identifiers of each category are as follows:

Factor 1: Hardware (factor classification identifier: 2-1);

Predictive indicators: whether the channel correction passes;

Factor classification identification: 2-1-1;

Factor 2: UE placement (factor classification identifier: 2-2);

Prediction indicator 1: RSRP balance between UE antennas;

Factor classification identification 1: 2-2-1;

Predictive indicator 2: UE placement position and method;

Factor classification identification 2: 2-2-2;

Factor three: base station RF (factor classification identification: 2-3);

Predictive indicator 1: direction angle (towards the building, increasing the reflection path);

Factor classification identification 1: 2-3-1;

Predictive indicator 2: Downtilt angle (open scenes increase ground reflection);

Factor classification identification 2: 2-3-2;

Factor 4: Algorithm (factor classification identifier: 2-4);

Predictive indicator 1: Tianxuan terminal: SRS right;

Factor classification identification 1: 2-4-1;

Predictive indicator 2: Unselected terminal: VAM+PMI right;

Factor classification identification 2: 2-4-2.

Step 2: Based on the random forest model, analyze the four scenario data of antenna coverage stored in the gNB log and the factor classification identifier in the Rank identifier, and obtain the most likely Rank failure factor in the scenario/sub-scenario;

Locate the area where the UE is located and generate an area identifier, and finally generate an identifier of the factor with the largest scene/sub-scenario Rank failure;

Store the area identifier, scene identifier and factor identifier into the Rank identifier (join the area identifier, scene identifier and factor identifier with the # symbol and put them into the Rank identifier, the format is: [region identifier]#[scene identifier]#[factor Identification]#[Factor classification identification]#Failure probability prediction value).

The second step is to build a random forest model and analyze the factors that have the greatest impact on improving Rank weight in common scenarios;

Combined with Figure 2, the specific description is as follows:

[Random Forest Model] Formula:

Parameter Description:

1. Set a constant n as the number of sampling scenes.

2. Where |Di|/|D| refers to the probability of scenarios one, two, three, and four. The total number brought in when calculating Hi is the number of scenarios one, two, three, and four. The Hi of each feature is obtained = the probability of abnormality in the scene.

For example: the historical log data brought into scenario one has |D| entries, and the abnormal data that matches scenario one (the AAU mechanical downtilt angle is not between 10 and 15°) has |Di| entries.

Random forest model running process:

1. First, the input is the sample set |D|;

2. Randomly select the training data set and sample features for |Di| round training;

2-1. Conduct the i-th random sampling of the training set, collecting n times in total, to obtain a sampling set containing n samples;

2-2. Use the sampling set |Di| to train the nth decision tree model Hi;

When training the node of the decision tree model, select a part of the sample features from all the sample features on the node, and select an optimal feature from these randomly selected part of the sample features to make the left and right subtree division results of the decision tree Hi;

3. Hj is equal to the weighted average of all probability predictions Hi of the scene, and the weighted average of the abnormality probability of the scene.

AAU, Active Antenna Unit, active antenna unit. AAU is the main equipment of 5G base station and an implementation of large-scale antenna array. AAU can be regarded as a combination of RRU and antenna, integrating multiple T/R units. The T/R unit is a radio frequency transceiver unit, which was first used in military phased array radars.

Common scenarios include:

Scenario 1: The wireless environment is single and there are few buildings.

Main scene identification: 1-1

Solution: The AAU mechanical downtilt angle is 10 to 15°. Try to cover the building reflection surface with the upper beam and the road with the lower beam. This will make it easier to produce multipath and increase the speed.

Second scene: The road is narrow and there are clusters of buildings on both sides.

Main scene identification: 1-2

Solution: AAU mechanical downtilt angle of 10° + narrow beam, antenna position aligned with the optimal reflecting surface of the building, allowing the beam signal to reflect back and forth between groups of buildings, creating a good multipath environment, improving Rank and speed .

Scene 3: Multi-lane intersection with clusters of buildings and wide road space.

Main scene identification: 1-3

Solution: Try to choose the optimal reflecting surface of the buildings on both sides of the intersection in the AAU coverage direction. Do not cover along the road. Create multi-faceted reflections as much as possible to improve the Rank.

Scene 4: Multi-lane road, single rows of buildings, shady trees.

Scene ID: 1-4

Solution: Select a single row of buildings or the ground for the AAU coverage direction, and create multipath through reflections from buildings, ground, and traffic flow, thereby improving Rank and speed.

Step 3: Based on the 5G Rank problem optimization measures and Rank identification analysis, build a Rank optimization program and associate it with the antenna coverage scenario/sub-scenario.

According to the above, the Rank identification format is: [Region identification]#[Scenario identification]#[Factor identification]#[Factor classification identification]#Failure probability prediction value.

In step three, when optimizing the downlink rate during the drive test, when an alarm occurs in the Rank link, the processing details are as follows:

S1: Split the Rank identifier to obtain the area where the UE is located.

S2, split the Rank identifier to obtain the corresponding scene/sub-scene.

S3: Extract the fault keyword based on the abnormal content of the drive test data report and compare it with the factor identifier obtained by splitting the Rank identifier. If the fault keyword appears one or more times in the abnormal content. This confirms that this exception is real and valid.

[Rank Optimizer] Detailed description:

Based on the analysis of Rank abnormal data in historical drive test abnormal log data, optimization measures for problems affecting Rank are obtained, and corresponding program presets are made based on commonly used optimization methods.

The 5G Rank problem optimization measures and Rank identification analysis are shown in Table 1.

Table 1 5G Rank problem optimization measures and Rank identification analysis table

The abbreviations and key terms used in this invention are defined as follows:

Rank: Number of spatial division multiplexing streams. A simple understanding is that the same time-frequency resources are divided into several parts in space and transmitted simultaneously. Codewords are mapped to each stream through layer mapping (number of codewords ≤ number of streams ≤ number of antenna ports). When time and frequency resources remain unchanged, the higher the Rank, the higher the actual throughput rate.

Principle and calculation method of Rank Rank principle In the field of communications, spatial multiplexing technology refers to sending different data on different antennas, also called spatial multiplexing. The standard for measuring spatial multiplexing is to look at the maximum number of different data that a system can send at each moment, which is called the "degree of freedom", that is, Rank. The larger the Rank, the greater the multiplexing gain. Codewords are mapped to each stream through layer mapping. The more layers, the higher the rate, and Rank determines the number of layers.

UE: User Equipment (UserEquipment), that is, mobile communication terminal equipment, such as mobile phones.

When a UE accesses the mobile communication network, the base station and the core network need to allocate resources to the UE. The resources can be divided from the following two perspectives: Network resources: such as the air interface channel resources and GTPU transmission resources allocated by the base station to the UE, and the core network allocation to the UE. GTPU transmission resources; system resources: such as threads, processes, single boards, virtual machines and other computer system resources used by base stations or core network equipment to serve UEs. The 3GPP mobile communication standard defines a RESET process between the base station and the core network. The purpose of this process is to notify the base station through the RESET process to initialize the system resources occupied by the UE and release the system resources occupied by the UE when a system resource failure in the core network fails and affects the UE. Network resources related to the UE; conversely, when a system resource failure in the base station affects the UE, the core network is notified through the RESET process to initialize the system resources occupied by the UE and release the network resources related to the UE.

The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above-mentioned embodiments. All technical solutions that fall under the idea of the present invention belong to the protection scope of the present invention. It should be pointed out that for those of ordinary skill in the art, several improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

A method for intelligent Rank downlink rate optimization that can be used for 6G, which is characterized by including:

Step 1: Combine the log data to classify the influencing factors of Rank. Based on the regional prediction model, obtain the predicted value of the failure probability of the road test Rank downlink rate, obtain the classification identifier of the factors that affect the Rank, and store the factor classification identifier and the predicted value of the failure probability. Rank identification;

Step 2: Based on the random forest model, analyze the four scenario data of antenna coverage stored in the gNB log and the factor classification identifier in the Rank identifier, to obtain the most likely Rank failure factors in the scenario/sub-scenario, and at the same time analyze the UE Locate the area and generate the area identifier, and finally generate the factor identifier with the largest scene/sub-scene Rank failure, and store the area identifier, scene identifier and factor identifier into the Rank identifier;

Step 3: Based on the 5G Rank problem optimization measures and Rank identification analysis, build a Rank optimization program and associate it with the antenna coverage scenario/sub-scenario to realize the self-healing adjustment and adjustment after the 5G Rank link fails during the UE's drive test downlink rate optimization process. optimization.
A method for intelligent Rank downlink rate optimization that can be used for 6G according to claim 1, characterized in that the step analyzes historical log data stored on the base station equipment to find out the data of factors affecting Rank and Classified according to hardware, UE placement, base station RF, and algorithm;

The most important factor data in each category is input into the constructed regional prediction model for analysis, and the sub-category failure probability prediction indicators of each category of factors that affect Rank are obtained, and the corresponding factor classification identification is generated.
A method for 6G intelligent Rank downlink rate optimization according to claim 2, characterized in that the regional prediction model in step one is:

P(A|B)=(P(B|A)*P(A))/P(B|A)P(A)+P(B|A')P(A')

P(A|B) is the probability of abnormality in factors affecting Rank;

P(B|A) is the probability of the result of the total number of abnormal items of factors affecting Rank/the total number of historical database items during the continuous learning process of the regional prediction model;

P(A) is the total number of data anomalies/the total number of historical data that ignores other factors and factors that affect Rank;

P(B|A') is the probability that the total number of data anomalies that affect Rank has appeared in the historical database;

P(A')=1-P(A).
A method for 6G intelligent Rank downlink rate optimization according to claim 3, characterized in that in the step one, the sub-category failure probability prediction indicators and corresponding factor classification identifiers of each category are as follows:

Factor 1: Hardware, corresponding factor classification identification: 2-1;

Prediction index: whether the channel correction is passed, corresponding factor classification identification: 2-1-1;

Factor 2: UE placement, corresponding factor classification identification: 2-2;

Prediction indicator 1: RSRP balance between UE antennas, corresponding factor classification identification 1: 2-2-1;

Prediction indicator 2: UE placement position and method, corresponding factor classification identification 2: 2-2-2;

Factor three: base station RF, corresponding factor classification identification: 2-3;

Predictive indicator 1: Toward the building, the direction angle of the reflection path is increased, corresponding to factor classification identification 1: 2-3-1;

Prediction indicator 2: Open scenes increase the downtilt angle of ground reflection, corresponding to factor classification identification 2: 2-3-2;

Factor 4: Algorithm, corresponding factor classification identifier: 2-4;

Prediction indicator 1: Tianxuan terminal: SRS right, corresponding factor classification identification 1: 2-4-1;

Prediction indicator 2: Unselected terminal: VAM+PMI right, corresponding factor classification identification 2: 2-4-2.
A method for intelligent Rank downlink rate optimization that can be used for 6G according to claim 1, characterized in that the first step is to store the factor classification identifier and the failure probability prediction value in the Rank identifier using a combination of # symbols, in the format It is: Factor classification identification #failure probability prediction value.
A method for 6G intelligent Rank downlink rate optimization according to claim 1, characterized in that the step two constructs a random forest model as follows:

Among them, n is the number of sampling scenes, which is 4, indicating that the sampling scenes include scenes one, two, three, and four;

|Di|/|D| refers to the probability of scenarios one, two, three and four;

H(i) is the total number of features in scenes one, two, three and four.
A method for 6G intelligent Rank downlink rate optimization according to claim 6, characterized in that the scenarios one, two, three and four and their scenario identifiers are:

Scenario 1: The wireless environment is single and there are few buildings. The main scene identification is: 1-1;

The second scene: the road is narrow, with buildings on both sides in groups, the main scene logo: 1-2;

Scene 3: Multi-lane intersection, clusters of buildings, open road space, main scene identification: 1-3;

Scene 4: Multi-lane road, single rows of buildings, shady trees, scene identification: 1-4.
A method for intelligent Rank downlink rate optimization that can be used for 6G according to claim 1, characterized in that the step two combines the area identifier, scene identifier, and factor identifier with the # symbol and puts them into the Rank identifier, The format is: area identification#scenario identification#factor identification#factor classification identification#failure probability prediction value.
A method for intelligent Rank downlink rate optimization that can be used for 6G according to claim 1, characterized in that the step three is performing drive test downlink rate optimization, and when an alarm occurs in the Rank link, the processing is as follows:

S1, split the Rank identifier to obtain the area where the UE is located;

S2, split the Rank identifier to obtain the corresponding scene/sub-scene;

S3: Extract the fault keyword based on the abnormal content of the drive test data report and compare it with the factor identifier obtained by splitting the Rank identifier. If the fault keyword appears in the abnormal content one or more times, it is confirmed that the abnormality is real and valid;

S4: Split the Rank identifier to obtain the predicted value of the probability of failure. If the predicted value of the probability is greater than 50%, execute the corresponding preset Rank optimization program to achieve self-healing or optimization of the fault.