Method for analyzing and restoring operation state of 5G-R wireless network

Abstract

The invention discloses a 5G-R wireless network operation state analysis and restoration method, and the method comprises the steps: collecting initial network data, carrying out the preprocessing of the initial network data, dividing the preprocessed initial network data into a training data set and a detection data set, and enabling the initial network data to comprise TCE data and network performance data; establishing a plurality of index systems based on data source types of the training data set, constructing an index model based on the plurality of index systems, and optimizing the index model to obtain a wireless network state analysis model; generating a wireless network operation state analysis result based on the wireless network state analysis model and the detection data set; and when the analysis result of the running state of the wireless network is that the network is abnormal, restoring the abnormal state of the wireless network in a reverse thrust manner, generating a restoration result, and establishing a wireless network state analysis model through TCE data and network performance data, thereby realizing accurate analysis and restoration of the running state of the 5G-R wireless network.

Description

Technical Field

[0001] The present application relates to the field of wireless network technology, and more specifically, to a method for analyzing and restoring the operating status of a 5G-R wireless network. Background Art

[0002] The GSM-R network has become one of the important communication means for the current high-speed rail transportation production command, and its healthy network status is an important guarantee for the stable operation of the railway. With the innovative application of new technologies, 5G technology, with its ultra-large bandwidth, ultra-low latency and massive connections, is continuously innovating its applications by integrating with vertical industries. 5G-R is a new generation of railway-specific mobile communication technology system innovatively developed on the basis of 5G technology and combined with railway application needs. It adapts to the needs of railway intelligent development and is in line with the development direction of international railway-specific mobile communications. Similarly, a healthy 5G-R wireless network status can ensure the normal operation of high-speed rail under 5G-R. Compared with the GSM-R network, 5G-R has major changes in network architecture, data types and collection methods. The existing diagnostic methods for the GSM-R wireless network status cannot be used to analyze the wireless network status of 5G-R.

[0003] Therefore, a better solution is urgently needed. Summary of the invention

[0004] The present invention provides a 5G-R wireless network operation status analysis and restoration method, which is used to solve the technical problem that the 5G-R wireless network operation status cannot be analyzed and restored in the prior art. The method includes:

[0005] Collecting initial network data, preprocessing the initial network data, and dividing the preprocessed initial network data into a training data set and a test data set, wherein the initial network data includes TCE data and network performance data;

[0006] Establishing multiple indicator systems based on the data source types of the training data set, constructing indicator models based on the multiple indicator systems, and optimizing the indicator models to obtain a wireless network status analysis model;

[0007] Generate a wireless network operation status analysis result based on the wireless network status analysis model and the detection data set;

[0008] When the wireless network operation status analysis result is that the network is abnormal, the abnormal state of the wireless network is restored by reverse inference, and a restoration result is generated.

[0009] In some specific embodiments, initial network data is collected, and the initial network data is preprocessed, and the preprocessed initial network data is divided into a training data set and a detection data set, specifically:

[0010] Acquire the initial network data through a northbound interface in the 5G-R network;

[0011] Performing de-duplication, de-noising and normalization processing on the initial network data to obtain standardized network data;

[0012] Dividing the canonical network data into a training data set and a test data set;

[0013] The training data set includes at least full data under normal network conditions for a preset number of days and sample data under various abnormal network conditions, and the preset number of days is no less than 3 days.

[0014] In some specific embodiments, multiple indicator systems are established based on the data source type of the training data set, and the TCE data includes MDT data, measurement report data, and signaling data, specifically:

[0015] When the data source type is MDT data, a first indicator system is established based on the MDT data, and the first indicator system includes the following indicators: UE idle state, UE connection state, radio link failure event, connection establishment failure event, switching failure time, access contention statistics, positioning information, sensor information, UE transmit power margin, DRB corresponding Qos statistics, PDCP layer uplink data flow, PDCP layer downlink data flow, uplink tail-removed packet flow, downlink tail-removed packet flow, downlink PDCP delay, uplink PDCP packet loss number, uplink PDCP successfully sent packet number, downlink PDCP packet loss number, downlink PDCP successfully sent packet number, and MDT data of abnormal terminals;

[0016] When the data source type is measurement report data, a second indicator system is established based on the measurement report data, and the second indicator system includes the following indicators: UE uplink packet loss rate, UE downlink packet loss rate, signal reception power of the serving cell, UE maximum transmit power of the serving cell, signal reception quality of the serving cell, time advance of the serving cell, PLMN of the UE accessing the NR serving cell, signal signal-to-noise ratio of the serving cell, UE transmit power of the serving cell, SSB reference signal reception power of NR cells with defined neighboring relationships and undefined neighboring relationships, SSB reference signal reception quality of NR cells with defined neighboring relationships and undefined neighboring relationships, and poor quality of the serving cell;

[0017] When the data source type is the signaling data, a third indicator system is established based on the signaling data, and the third indicator system includes the following indicators: registration management, deregistration management, PDU session creation, UE context management, UE mobility management, paging process, NAS message transmission process, Trace process, location reporting process, UE wireless capability management process, switching preparation process;

[0018] When the data source type is network performance data, a fourth indicator system is established based on the network performance data, and the fourth indicator system includes the following indicators: wireless bearer establishment success rate, wireless resource control connection establishment success rate, 5G independent networking access success rate, context activation success rate, wireless bearer abnormal release rate, gNB base station CPU average load, gNB base station CPU peak load, average number of RRC connected users, total number of uplink physical resource blocks occupied in the cell, total number of uplink PRBs available in the cell, total number of downlink PRBs occupied in the cell, total number of downlink PRBs available in the cell, average downlink packet delay, average uplink rate, average downlink rate, drop rate, and switching success rate.

[0019] In some specific embodiments, an indicator model is constructed based on the multiple indicator systems, specifically:

[0020] The indicators in the first indicator system, the second indicator system, the third indicator system and the fourth indicator system are divided into numerical indicators and non-numerical indicators;

[0021] Constructing a first indicator model based on the numerical type indicator and the first preset training rule;

[0022] A second indicator model is constructed based on the non-numerical type indicator and the second preset training rule.

[0023] In some specific embodiments, the first preset training rule includes N first judgment units, where N is greater than or equal to 1, and the number of the first judgment units is consistent with the number of associated features of the input numerical type indicator. In each first judgment unit, a first logical judgment is performed on each associated feature, and after the first logical judgment of each first judgment unit passes, it is determined that the input numerical type indicator matches the first indicator model, and the model result is recorded. If the first logical judgment of any first judgment unit fails, the match is unsuccessful.

[0024] In some specific embodiments, the second preset training rule includes N+1 second judgment units, where N is greater than or equal to 1, and the number of the second judgment units is the number of associated features of the input non-numeric type indicator plus 1, and the N+1th judgment unit is a logical sequence judgment of the first second judgment unit to the Nth judgment unit, and a second logical judgment is performed on each associated feature in each second judgment unit, and after the second logical judgment of each second judgment unit passes and the logical sequence judgment passes, it is determined that the input numerical type indicator matches the second indicator model, and the model result is recorded. If the first logical judgment of any judgment unit fails or the logical sequence judgment fails, the match is unsuccessful.

[0025] In some specific embodiments, the indicator model is optimized to obtain a wireless network status analysis model, including:

[0026] In the first indicator model, the indicator baseline of the first indicator model and the baseline of each indicator are optimized by using the full data, and verified based on the sample data to obtain an optimized first indicator model;

[0027] In the second indicator model, the characteristic value relationship of the second indicator model is adjusted by the full data, and verified based on the sample data to obtain an optimized second indicator model;

[0028] The optimized first indicator model and the optimized second indicator model are used as the wireless network status analysis model.

[0029] In some specific embodiments, in the first indicator model, the indicator baseline of the first indicator model and the baseline of each indicator are optimized through the full data, and verified based on the sample data to obtain the optimized first indicator model, which is specifically:

[0030] In the first indicator model, the indicator baseline of the first indicator function is trained based on the full data, and the trained indicator baseline value is verified based on the sample data;

[0031] In the first indicator model, an indicator data range of each indicator is obtained, and a data distribution set arranged in sequence is generated based on the indicator data range;

[0032] When the indicator value is a positive indicator, a first ratio in the data distribution set that is greater than a preset baseline corresponding to each indicator is obtained, and when the first ratio is greater than a first preset value, verification is performed based on the sample data, and when the first ratio is less than or equal to the first preset value, the first preset value is increased by one, and then it is re-determined whether the first ratio is greater than the adjusted first preset value, until the first ratio is greater than the first preset value;

[0033] When the indicator value is a negative indicator, obtain the second proportion of the data distribution set that is smaller than the preset baseline corresponding to each indicator, and when the second proportion is greater than the second preset value, verify based on the sample data, and when the second proportion is less than or equal to the second preset value, add one to the second preset value, and re-judge whether the second proportion is greater than the adjusted second preset value, until the second proportion is greater than the second preset value.

[0034] In some specific embodiments, a wireless network operation status analysis result is generated based on the wireless network status analysis model and the detection data set, and the wireless network status analysis model includes an indicator layer, an analysis layer and a conclusion layer, specifically:

[0035] Inputting the detection data set into the wireless network operation status analysis model, and obtaining an abnormality inspection set of each indicator in the detection data set based on the indicator layer;

[0036] By means of a preset expert database and statistical analysis, the abnormal inspection set is analyzed based on the analysis layer to obtain an analysis result;

[0037] The analysis result is input into the conclusion layer, and a binary method is used to obtain the wireless network operation status analysis result, wherein the wireless network operation status analysis result is specifically a healthy network or an abnormal network.

[0038] In some specific embodiments, when the wireless network operation status analysis result is a network abnormality, a reverse inference method is used to restore the abnormal state of the wireless network and generate a restoration result, specifically:

[0039] Obtaining the wireless network operation status analysis results in the conclusion layer, and extracting network status abnormality scenarios and network abnormality lists;

[0040] Acquiring analysis reasons in the analysis layer based on the network anomaly list;

[0041] Extracting corresponding abnormal indicators in the abnormal inspection set in the indicator layer based on the analysis reasons;

[0042] Key information in the abnormal indicator is obtained, and the abnormal state of the wireless network is restored based on the key information to obtain a restoration result.

[0043] By applying the above technical scheme, a method for analyzing and restoring the operation status of a 5G-R wireless network is proposed, the method comprising: collecting initial network data, preprocessing the initial network data, and dividing the preprocessed initial network data into a training data set and a detection data set, the initial network data comprising TCE data and network performance data; establishing multiple indicator systems based on the data source type of the training data set, constructing an indicator model based on the multiple indicator systems, and optimizing the indicator model to obtain a wireless network status analysis model; generating a wireless network operation status analysis result based on the wireless network status analysis model and the detection data set; when the wireless network operation status analysis result is a network abnormality, restoring the abnormal state of the wireless network by reverse inference, and generating a restoration result, establishing a wireless network status analysis model through TCE data and network performance data, and realizing accurate analysis and restoration of the operation status of the 5G-R wireless network. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 It is a flow chart of a 5G-R wireless network operation status analysis and restoration method provided in an embodiment of the present application;

[0046] Figure 2 It is a schematic diagram of the principle of the process of obtaining initial network data provided by an embodiment of the present application;

[0047] Figure 3 It is a flowchart of the construction process of the first indicator model provided in the embodiment of the present application;

[0048] Figure 4 It is a flowchart of the construction process of the second indicator model provided in the embodiment of the present application;

[0049] Figure 5 It is a schematic diagram of the principle of the wireless network operation status analysis process provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0051] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0052] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0053] First, the terms appearing in this application are explained:

[0054] TCE: trace collection entity (TCE)

[0055] MDT: Minimization of Drive Tests (MDT)

[0056] UE: User Equipment (UE)

[0057] DRB: data radio bearer (DRB)

[0058] Qos: Quality of Service (Qos)

[0059] PDCP: Packet Data Convergence Protocol (PDCP)

[0060] NR: New Radio Access Network (New RAN, NR)

[0061] PLMN: Public Land Mobile Network (PLMN)

[0062] SSB: Synchronization Signal and PBCH block (SSB)

[0063] PDU: Protocol Data Unit (PDU)

[0064] NAS: Non-Access Stratum (NAS)

[0065] Trace:

[0066] RRC: Radio Resource Control (RRC)

[0067] SA: Stand alone (SA)

[0068] PRB: physical resource block (PRB)

[0069] 5G-R: A new generation of railway mobile communication system based on 5G technology (5G-Railway, 5G-R)

[0070] GSM-R: Global System for Mobile Communications-Railway (GSM-R)

[0071] RLF: Radio link failure (RLF) event

[0072] CEF: Connection Establishment Failure (CEF) event (Connection Establishment Failure, CEF)

[0073] like Figure 1 As shown, the present application proposes a 5G-R wireless network operation status analysis and restoration method, the method comprising the following steps:

[0074] Step S101, collecting initial network data, preprocessing the initial network data, and dividing the preprocessed initial network data into a training data set and a detection data set, wherein the initial network data includes TCE data and network performance data.

[0075] In a possible implementation, initial network data is collected, preprocessed, and the preprocessed initial network data is divided into a training data set and a detection data set, specifically:

[0076] Acquire the initial network data through a northbound interface in the 5G-R network;

[0077] Performing de-duplication, de-noising and normalization processing on the initial network data to obtain standardized network data;

[0078] Dividing the canonical network data into a training data set and a test data set;

[0079] The training data set includes at least full data under normal network conditions for a preset number of days and sample data under various abnormal network conditions, and the preset number of days is no less than 3 days.

[0080] In this embodiment, the TCE data and network performance data in 5G-R can reflect the connection status of the wireless network, the quality of network-borne services, etc., so the initial network data selects TCE data and network performance data.

[0081] In this embodiment, Figure 2 As shown, the required initial network data is obtained through the newly established northbound interface of the 5G-R network. The gNodeB uploads the measurement report data obtained from the UE and stores it on the TCE server. The TCE provides a northbound interface for various data stored in it. The third party obtains the MDT data through the MDT docking interface on demand according to the MDT data authorization. Other data are similarly processed and will not be elaborated here.

[0082] In this embodiment, the collected data is cleaned and preprocessed by deduplication, denoising and normalization to obtain standardized data, which is recorded as a set DATA. The collected data is divided into a training data set Tr and a detection data set Te according to different functions. Tr is used when constructing a model. In order to make the model more versatile and applicable, Tr needs to include at least N (N>=3) days of full data Tr_normal under normal network conditions and sample data Tr_fault under various abnormal network conditions. Tr is only collected and used in the modeling stage. In other cases, the collected data is used as a detection data set.

[0083] Step S102: establishing a plurality of indicator systems based on the data source types of the training data set, constructing an indicator model based on the plurality of indicator systems, and optimizing the indicator model to obtain a wireless network status analysis model.

[0084] In a possible implementation, multiple indicator systems are established based on the data source type of the training data set, and the TCE data includes MDT data, measurement report data, and signaling data, specifically:

[0085] When the data source type is MDT data, a first indicator system is established based on the MDT data, and the first indicator system includes the following indicators: UE idle state, UE connection state, radio link failure event, connection establishment failure event, switching failure time, access contention statistics, positioning information, sensor information, UE transmit power margin, DRB corresponding Qos statistics, PDCP layer uplink data flow, PDCP layer downlink data flow, uplink tail-removed packet flow, downlink tail-removed packet flow, downlink PDCP delay, uplink PDCP packet loss number, uplink PDCP successfully sent packet number, downlink PDCP packet loss number, downlink PDCP successfully sent packet number, and MDT data of abnormal terminals;

[0086] When the data source type is measurement report data, a second indicator system is established based on the measurement report data, and the second indicator system includes the following indicators: UE uplink packet loss rate, UE downlink packet loss rate, signal reception power of the serving cell, UE maximum transmit power of the serving cell, signal reception quality of the serving cell, time advance of the serving cell, PLMN of the UE accessing the NR serving cell, signal signal-to-noise ratio of the serving cell, UE transmit power of the serving cell, SSB reference signal reception power of NR cells with defined neighboring relationships and undefined neighboring relationships, SSB reference signal reception quality of NR cells with defined neighboring relationships and undefined neighboring relationships, and poor quality of the serving cell;

[0087] When the data source type is the signaling data, a third indicator system is established based on the signaling data, and the third indicator system includes the following indicators: registration management, deregistration management, PDU session creation, UE context management, UE mobility management, paging process, NAS message transmission process, Trace process, location reporting process, UE wireless capability management process, switching preparation process;

[0088] When the data source type is network performance data, a fourth indicator system is established based on the network performance data, and the fourth indicator system includes the following indicators: wireless bearer establishment success rate, wireless resource control connection establishment success rate, 5G independent networking access success rate, context activation success rate, wireless bearer abnormal release rate, gNB base station CPU average load, gNB base station CPU peak load, average number of RRC connected users, total number of uplink physical resource blocks occupied in the cell, total number of uplink PRBs available in the cell, total number of downlink PRBs occupied in the cell, total number of downlink PRBs available in the cell, average downlink packet delay, average uplink rate, average downlink rate, drop rate, and switching success rate.

[0089] In this embodiment, different types of data sources are used, so different indicator systems are established according to different types of data. The indicator system is composed of a series of indicators, and each indicator is an item. The indicator system established for MDT data is ITEMS_MDT, including UE status, radio link failure (RLF) events, connection establishment failure (CEF) events, MDT of abnormal terminals, etc.; the indicator system established for measurement report data is ITEMS_MR, including UE uplink packet loss rate, signal reception quality of the serving cell, and service cell quality to 7, etc.; the indicator system established for signaling data is ITEMS_SI, including registration management, UE mobility management, paging process, switching preparation process, etc.; the indicator system established for network performance data is ITEMS_PF, including radio resource control (RRC) connection establishment success rate, 5G stand alone (SA) access success rate, downlink packet average delay, drop rate, etc.

[0090] The specific system composition is shown in Table 1 below.

[0091]

[0092]

[0093]

[0094] Table 1

[0095] In a possible implementation, an indicator model is constructed based on the multiple indicator systems, specifically:

[0096] The indicators in the first indicator system, the second indicator system, the third indicator system and the fourth indicator system are divided into numerical indicators and non-numerical indicators;

[0097] Constructing a first indicator model based on the numerical type indicator and the first preset training rule;

[0098] A second indicator model is constructed based on the non-numerical type indicator and the second preset training rule.

[0099] In one possible implementation, the first preset training rule includes N first judgment units, where N is greater than or equal to 1, and the number of the first judgment units is consistent with the number of associated features of the input numerical type indicator. A first logical judgment is performed on each associated feature in each first judgment unit, and after the first logical judgment of each first judgment unit passes, it is determined that the input numerical type indicator matches the first indicator model, and the model result is recorded. If the first logical judgment of any first judgment unit fails, the match is unsuccessful.

[0100] In one possible implementation, the second preset training rule includes N+1 second judgment units, where N is greater than or equal to 1, and the number of the second judgment units is the number of associated features of the input non-numeric type indicator plus 1, and the N+1th judgment unit is a logical sequence judgment of the first second judgment unit to the Nth judgment unit, and a second logical judgment is performed on each associated feature in each second judgment unit, and after the second logical judgment of each second judgment unit passes and the logical sequence judgment passes, it is determined that the input numerical type indicator matches the second indicator model, and the model result is recorded. If the first logical judgment of any judgment unit fails or the logical sequence judgment fails, the match is unsuccessful.

[0101] In this embodiment, the indicator items under the four indicator systems are analyzed, and the items can be divided into two types, which are established according to two indicator models. The first type is that the baseline value fields involved in the indicators are all numerical types, such as the service cell quality difference indicator in the measurement report indicator system, which requires the four fields of uplink level, downlink level, uplink quality, and downlink quality to be numerical types before the indicator can be judged. The switching success rate and call drop rate indicators in the performance indicator system can be represented only by the switching success rate and call drop rate field values ​​in the data. The second type is that the indicator requires one or more fields, and the fields are non-numerical types, such as the registration process indicator in the signaling indicator system, which requires multiple signaling data to exist at the same time and the signaling sequence to be correct to judge the indicator as normal. The lack of key signaling or abnormal sequence will judge the indicator as abnormal. Therefore, two indicator models are constructed to adapt to the indicators under four different indicator systems.

[0102] In this embodiment, the construction process of the first indicator model is as follows: Figure 3 As shown, the first indicator model includes N characteristic values ​​(N is greater than or equal to 1) including the first characteristic, the second characteristic, etc. Each characteristic corresponds to a judgment unit, and in each judgment unit, Figure 3 As shown in the figure, a preset threshold is set. When the associated features in the judgment unit meet the following conditions Figure 3 The first judgment unit can only be satisfied when the value in is greater than or equal to or less than the preset threshold and is a numerical value, and then the N features need to be satisfied at the same time to match the model. The baseline value (preset threshold) involved in the model is set according to the wireless network operation specification and the operator's work experience value.

[0103] In this embodiment, the construction process of the second indicator model is as follows: Figure 4 As shown, the second indicator model includes N characteristic values ​​(N is greater than or equal to 1) including the first characteristic, the second characteristic, etc., and the first characteristic, the second characteristic, etc. need to meet the logical order as the N+1th characteristic, and the model is matched if any characteristic value is met. The first N characteristic values ​​are set according to the normal service interaction of the wireless network in the wireless service specification.

[0104] In a possible implementation, the indicator model is optimized to obtain a wireless network status analysis model, including:

[0105] In the first indicator model, the indicator baseline of the first indicator model and the baseline of each indicator are optimized by using the full data, and verified based on the sample data to obtain an optimized first indicator model;

[0106] In the second indicator model, the characteristic value relationship of the second indicator model is adjusted by the full data, and verified based on the sample data to obtain an optimized second indicator model;

[0107] The optimized first indicator model and the optimized second indicator model are used as the wireless network status analysis model.

[0108] In a possible implementation, in the first indicator model, the indicator baseline and the baseline of each indicator of the first indicator model are optimized by the full data, and verified based on the sample data to obtain the optimized first indicator model, specifically:

[0109] In the first indicator model, the indicator baseline of the first indicator function is trained based on the full data, and the trained indicator baseline value is verified based on the sample data;

[0110] In the first indicator model, an indicator data range of each indicator is obtained, and a data distribution set arranged in sequence is generated based on the indicator data range;

[0111] When the indicator value is a positive indicator, a first ratio in the data distribution set that is greater than a preset baseline corresponding to each indicator is obtained, and when the first ratio is greater than a first preset value, verification is performed based on the sample data, and when the first ratio is less than or equal to the first preset value, the first preset value is increased by one, and then it is re-determined whether the first ratio is greater than the adjusted first preset value, until the first ratio is greater than the first preset value;

[0112] When the indicator value is a negative indicator, obtain the second proportion of the data distribution set that is smaller than the preset baseline corresponding to each indicator, and when the second proportion is greater than the second preset value, verify based on the sample data, and when the second proportion is less than or equal to the second preset value, add one to the second preset value, and re-judge whether the second proportion is greater than the adjusted second preset value, until the second proportion is greater than the second preset value.

[0113] In this embodiment, after the model is constructed, the cross-validation method is used to perform model validation optimization. For the first indicator model, the training indicator baseline value is obtained through the data set Tr_normal, and then the rationality of the model indicator baseline value is verified according to Tr_fault. For a specific indicator, the indicator data range range = [r1, rn] in Tr_normal is calculated, and the data distribution set collection = {(r1, n1), (r2, n2), (r3, n3) ... (rn, nn)} arranged in order, where rn is the nth value in the data set, and nn is the number of statistics of the nth value rn. The proportion of the preset baseline value base of the calculated indicator is greater than S%, that is, when the indicator value is a positive indicator such as the success rate, the proportion of data greater than base in the collection is calculated, and when the indicator value is a negative indicator such as the call drop rate, the proportion of data less than base in the collection is calculated. If the proportion is greater than S%, the baseline value base is set reasonably. If it is not satisfied, then fine-tune base ± 1 and recalculate until it meets the conditions. Then, the model's ability to detect abnormal indicator values ​​is verified based on Tr_fault. If all abnormal indicators in Tr_fault can be detected according to the set base value, it means that the model setting is reasonable. If some abnormal sample data is not detected, the indicator baseline value is corrected according to the abnormal sample data, and continuous testing and improvement are carried out until all abnormal indicators are detected.

[0114] For the second indicator model, according to Tr_normal and Tr_fault, the eigenvalue relationship of the model is improved. For example, the first feature signaling and the second feature signaling are adjusted from the original parallel relationship to an or relationship according to the training set data, or the Nth feature signaling has no effect. The model is continuously improved to avoid detecting normal data as abnormal, resulting in an increase in network status abnormalities and false alarms, or filtering out abnormal data, resulting in network status abnormalities not being reported.

[0115] Step S103: generating a wireless network operation status analysis result based on the wireless network status analysis model and the detection data set.

[0116] In a possible implementation, a wireless network operation status analysis result is generated based on the wireless network status analysis model and the detection data set, wherein the wireless network status analysis model includes an indicator layer, an analysis layer, and a conclusion layer, specifically:

[0117] Inputting the detection data set into the wireless network operation status analysis model, and obtaining an abnormality inspection set of each indicator in the detection data set based on the indicator layer;

[0118] By means of a preset expert database and statistical analysis, the abnormal inspection set is analyzed based on the analysis layer to obtain an analysis result;

[0119] The analysis result is input into the conclusion layer, and a binary method is used to obtain the wireless network operation status analysis result, wherein the wireless network operation status analysis result is specifically a healthy network or an abnormal network.

[0120] In this embodiment, Figure 5 As shown, based on the constructed indicator model, the wireless network status is analyzed at three levels: indicator layer, analysis layer, and conclusion layer. The indicator layer imports the detection data set Te into the indicator model to obtain the abnormal detection set exception_items of each indicator. Based on the previous direct conclusion of network status through indicator layer detection, this patent adds an analysis layer, which can deeply analyze the correlation between indicators and diagnose the cause of network abnormality. The analysis layer analyzes the exception_items set and uses two methods to assist each other in analysis. One is to analyze the cause of network abnormality based on the knowledge of the expert library. The second is to analyze and predict by combining historical data and indicators in a statistical way. The conclusion layer draws a conclusion of network health or network abnormality based on the dichotomy method for the causes obtained by the analysis layer.

[0121] In this embodiment, the analysis layer adopts two analysis methods. The first method is to analyze the abnormal reasons of indicators based on the expert knowledge base, in a similar form to {item1, item2}→reason1, {item1, item3}→reason2, {item2∪item3, item4}→reason3, and different indicator combinations are matched to different reasons according to the knowledge base. For example, when judging the existence of interference source in the service cell through the expert knowledge base, the indicator one, poor service cell quality, is judged. If it exists, it passes, and if it does not exist, it passes the indicator two, call drop rate. If it exists, it passes, and if it does not exist, it does not meet the requirements. The reasons for network abnormality analysis are gradually obtained, which are recorded as the exception_reasons set.

[0122] The second method is based on statistical analysis. Correlation analysis is conducted from the following dimensions:

[0123] Statistical indicator abnormal distribution: Based on the current abnormal indicator concentration, the number of abnormal indicators is counted according to the cell or kilometer range to determine the existence of interference.

[0124] Statistical indicator occurrence time: Statistics are conducted by time period to check whether it is time-related, whether there are occasional interferences or regular problems.

[0125] Statistical terminal: Collect statistics based on the abnormal terminal indicators to check whether the abnormality is caused by hardware equipment, etc.

[0126] Statistical abnormal change trend: For indicators that have not experienced abnormalities this time, combined with historical data, the probability of abnormalities occurring in the future can be predicted based on the trends and patterns of historical data.

[0127] The conclusion layer uses a binary classification method to obtain the conclusion result of network health and network anomaly through the cause of the analysis layer. Through the expert knowledge base, statistical indicator abnormal distribution, statistical indicator occurrence time, and statistical terminal, the analysis is based on abnormal indicators, corresponding to the network abnormality conclusion, and the obtained result set is summarized and put into the abnormal list exception_list. The statistical abnormal change trend is based on the abnormal indicator this time, based on the historical data analysis, there may be an abnormal probability, corresponding to the network health conclusion, and the obtained result set is put into the potential list warning_list.

[0128] Step S104: when the wireless network operation status analysis result is that the network is abnormal, the abnormal state of the wireless network is restored by reverse inference, and a restoration result is generated.

[0129] In a possible implementation, when the wireless network operation status analysis result is a network abnormality, a reverse inference method is used to restore the abnormal state of the wireless network and generate a restoration result, specifically:

[0130] Obtaining the wireless network operation status analysis results in the conclusion layer, and extracting network status abnormality scenarios and network abnormality lists;

[0131] Acquiring analysis reasons in the analysis layer based on the network anomaly list;

[0132] Extracting corresponding abnormal indicators in the abnormal inspection set in the indicator layer based on the analysis reasons;

[0133] Key information in the abnormal indicator is obtained, and the abnormal state of the wireless network is restored based on the key information to obtain a restoration result.

[0134] In this embodiment, the overall situation of the wireless network can be understood by analyzing the operation status of the wireless network. A further requirement is to be able to regulate the network in a targeted manner. Therefore, after analyzing the operation status of the wireless network, the present invention restores the network abnormal state process, locates the network abnormal points, and improves the efficiency of network optimization.

[0135] The restoration of the abnormal state of the wireless network adopts the reverse inference method of the analysis process, and performs the restoration process from the following levels. The first layer is the network state analysis conclusion layer. According to the conclusion layer in the previous step, the wireless network operation state analysis conclusion result is obtained, and the network state abnormality scenario and the network abnormality list exception_list are extracted. The second layer is the abnormal analysis reason layer. According to the network abnormality list, the analysis layer in the previous step is extracted to obtain the analysis reason exception_reasons. The third layer is the abnormal indicator layer. According to the exception_reasons in the abnormal indicator set exception_items, the corresponding abnormal indicator is proposed. The fourth layer is the abnormal data layer. The data corresponding to the abnormal indicator is proposed in the DATA set. The data includes key information such as time, community, and train number.

[0136] In this embodiment, there is no abnormality in the potential list, but there may be problems with the historical trend deduction, and the state evolution is restored in the form of a trend chart.

[0137] In summary, by analyzing the wireless network status based on TCE data and network performance data, maintenance personnel in the work area can understand the wireless network status in real time, handle problems immediately, clearly see the causes of wireless network anomalies, and accurately issue network optimization work orders. Network optimization personnel can narrow the scope of investigation and improve work efficiency based on the location of wireless network anomalies. For management personnel, it can help understand the overall wireless network status and trends, and formulate wireless network-related work in a targeted manner.

[0138] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0139] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0140] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0141] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. A 5G-R wireless network operation status analysis and restoration method, characterized in that: The method comprises: Collecting initial network data, preprocessing the initial network data, and dividing the preprocessed initial network data into a training data set and a test data set, wherein the initial network data includes TCE data and network performance data; Establishing multiple indicator systems based on the data source types of the training data set, constructing indicator models based on the multiple indicator systems, and optimizing the indicator models to obtain a wireless network status analysis model; Generate a wireless network operation status analysis result based on the wireless network status analysis model and the detection data set; When the wireless network operation status analysis result is that the network is abnormal, the abnormal state of the wireless network is restored by reverse inference, and a restoration result is generated.

2. The method according to claim 1, characterized in that Collect initial network data, preprocess the initial network data, and divide the preprocessed initial network data into a training data set and a test data set, specifically: Acquire the initial network data through a northbound interface in the 5G-R network; Performing de-duplication, de-noising and normalization processing on the initial network data to obtain standardized network data; Dividing the canonical network data into a training data set and a test data set; The training data set includes at least full data under normal network conditions for a preset number of days and sample data under various abnormal network conditions, and the preset number of days is no less than 3 days.

3. The method according to claim 2, characterized in that Multiple indicator systems are established based on the data source types of the training data set. The TCE data includes MDT data, measurement report data, and signaling data, specifically: When the data source type is MDT data, a first indicator system is established based on the MDT data, and the first indicator system includes the following indicators: UE idle state, UE connection state, radio link failure event, connection establishment failure event, switching failure time, access contention statistics, positioning information, sensor information, UE transmit power margin, DRB corresponding Qos statistics, PDCP layer uplink data flow, PDCP layer downlink data flow, uplink tail-removed packet flow, downlink tail-removed packet flow, downlink PDCP delay, uplink PDCP packet loss number, uplink PDCP successfully sent packet number, downlink PDCP packet loss number, downlink PDCP successfully sent packet number, and MDT data of abnormal terminals; When the data source type is measurement report data, a second indicator system is established based on the measurement report data, and the second indicator system includes the following indicators: UE uplink packet loss rate, UE downlink packet loss rate, signal reception power of the serving cell, UE maximum transmit power of the serving cell, signal reception quality of the serving cell, time advance of the serving cell, PLMN of the UE accessing the NR serving cell, signal signal-to-noise ratio of the serving cell, UE transmit power of the serving cell, SSB reference signal reception power of NR cells with defined neighboring relationships and undefined neighboring relationships, SSB reference signal reception quality of NR cells with defined neighboring relationships and undefined neighboring relationships, and poor quality of the serving cell; When the data source type is the signaling data, a third indicator system is established based on the signaling data, and the third indicator system includes the following indicators: registration management, deregistration management, PDU session creation, UE context management, UE mobility management, paging process, NAS message transmission process, Trace process, location reporting process, UE wireless capability management process, switching preparation process; When the data source type is network performance data, a fourth indicator system is established based on the network performance data, and the fourth indicator system includes the following indicators: wireless bearer establishment success rate, wireless resource control connection establishment success rate, 5G independent networking access success rate, context activation success rate, wireless bearer abnormal release rate, gNB base station CPU average load, gNB base station CPU peak load, average number of RRC connected users, total number of uplink physical resource blocks occupied in the cell, total number of uplink PRBs available in the cell, total number of downlink PRBs occupied in the cell, total number of downlink PRBs available in the cell, average downlink packet delay, average uplink rate, average downlink rate, drop rate, and switching success rate.

4. The method according to claim 3, characterized in that An indicator model is constructed based on the multiple indicator systems, specifically: The indicators in the first indicator system, the second indicator system, the third indicator system and the fourth indicator system are divided into numerical indicators and non-numerical indicators; Constructing a first indicator model based on the numerical type indicator and the first preset training rule; A second indicator model is constructed based on the non-numerical type indicator and the second preset training rule.

5. The method according to claim 4, characterized in that The first preset training rule includes N first judgment units, where N is greater than or equal to 1, and the number of the first judgment units is consistent with the number of associated features of the input numerical type indicator. A first logical judgment is performed on each associated feature in each first judgment unit, and after the first logical judgment of each first judgment unit passes, it is determined that the input numerical type indicator matches the first indicator model, and the model result is recorded. If the first logical judgment of any first judgment unit fails, the match is unsuccessful.

6. The method according to claim 4, characterized in that The second preset training rule includes N+1 second judgment units, where N is greater than or equal to 1, and the number of the second judgment units is the number of associated features of the input non-numeric type indicator plus 1, and the N+1th judgment unit is a logical sequence judgment of the first second judgment unit to the Nth judgment unit, and a second logical judgment is performed on each associated feature in each second judgment unit, and after the second logical judgment of each second judgment unit passes and the logical sequence judgment passes, it is determined that the input numerical type indicator matches the second indicator model, and the model result is recorded. If the first logical judgment of any judgment unit fails or the logical sequence judgment fails, the match is unsuccessful.

7. The method according to claim 6, characterized in that The indicator model is optimized to obtain a wireless network status analysis model, including: In the first indicator model, the indicator baseline of the first indicator model and the baseline of each indicator are optimized by using the full data, and verified based on the sample data to obtain an optimized first indicator model; In the second indicator model, the characteristic value relationship of the second indicator model is adjusted by the full data, and verified based on the sample data to obtain an optimized second indicator model; The optimized first indicator model and the optimized second indicator model are used as the wireless network status analysis model.

8. The method according to claim 7, characterized in that In the first indicator model, the indicator baseline of the first indicator model and the baseline of each indicator are optimized through the full data, and verified based on the sample data to obtain the optimized first indicator model, which is specifically: In the first indicator model, the indicator baseline of the first indicator function is trained based on the full data, and the trained indicator baseline value is verified based on the sample data; In the first indicator model, an indicator data range of each indicator is obtained, and a data distribution set arranged in sequence is generated based on the indicator data range; When the indicator value is a positive indicator, a first ratio in the data distribution set that is greater than a preset baseline corresponding to each indicator is obtained, and when the first ratio is greater than a first preset value, verification is performed based on the sample data, and when the first ratio is less than or equal to the first preset value, the first preset value is increased by one, and then it is re-determined whether the first ratio is greater than the adjusted first preset value, until the first ratio is greater than the first preset value; When the indicator value is a negative indicator, obtain the second proportion of the data distribution set that is smaller than the preset baseline corresponding to each indicator, and when the second proportion is greater than the second preset value, verify based on the sample data, and when the second proportion is less than or equal to the second preset value, add one to the second preset value, and re-judge whether the second proportion is greater than the adjusted second preset value, until the second proportion is greater than the second preset value.

9. The method according to claim 8, characterized in that Generate a wireless network operation status analysis result based on the wireless network status analysis model and the detection data set, wherein the wireless network status analysis model includes an indicator layer, an analysis layer and a conclusion layer, specifically: Inputting the detection data set into the wireless network operation status analysis model, and obtaining an abnormality inspection set of each indicator in the detection data set based on the indicator layer; By means of a preset expert database and statistical analysis, the abnormal inspection set is analyzed based on the analysis layer to obtain an analysis result; The analysis result is input into the conclusion layer, and a binary method is used to obtain the wireless network operation status analysis result, wherein the wireless network operation status analysis result is specifically a healthy network or an abnormal network.

10. The method according to claim 1, characterized in that When the wireless network operation status analysis result is that the network is abnormal, the abnormal state of the wireless network is restored by reverse inference, and a restoration result is generated, specifically: Obtaining the wireless network operation status analysis results in the conclusion layer, and extracting network status abnormality scenarios and network abnormality lists; Acquiring analysis reasons in the analysis layer based on the network anomaly list; Extracting corresponding abnormal indicators in the abnormal inspection set in the indicator layer based on the analysis reasons; Key information in the abnormal indicator is obtained, and the abnormal state of the wireless network is restored based on the key information to obtain a restoration result.

Images