CN114301767A - Optical communication network robustness analysis method based on interpretation degree and high survivability algorithm - Google Patents

Abstract

The invention discloses an optical communication network robustness analysis method based on interpretation degree and high survivability algorithm, which comprises the following steps: 1) node fault positioning; 2) and (5) ensuring the survivability of the nodes. The method adopts the probability model to express the relation between the network node fault and the symptom, so that the optimal fault hypothesis set is simply and effectively judged, and the troubleshooting of the fault and the judgment on the network robustness can be more accurate; and a survivability high-survivability anti-damage algorithm is added to further judge the anti-damage capability of the network, so that the robustness of the whole network is effectively analyzed and is more reliable.

Description

Optical communication network robustness analysis method based on interpretation degree and high survivability algorithm

Technical Field

The invention relates to the field of optical communication, in particular to an optical communication network robustness analysis method based on interpretation degree and high survivability algorithm in an optical communication network.

Background

Optical network communication is a communication method using light waves as information carriers and optical fibers as transmission media, and a network structure formed in this way can be called an optical communication network. Since the concept of optical fiber communication proposed by high-roll in 1966, optical communication networks have been extensively studied by many scholars and widely used in real life, and have become the main support of modern communication.

The method aims at solving the problem that in the current practical optical communication network engineering application, no matter new construction, transformation or upgrading, reliable planning and design of the network are needed, especially the method is suitable for the urgent needs of large-scale complex network construction planning, engineering construction, network evaluation, network optimization and the like, and the traditional method faces the bottleneck problem depending on empirical calculation, so that the relevant reliability analysis such as network robustness, survivability, fault simulation and the like is difficult to realize. The invention considers the service characteristics and demand factors of the optical network, establishes a network robustness evaluation method, integrates network nodes and a service fault analysis method, and realizes effective analysis of the network robustness.

Disclosure of Invention

The invention aims to provide an optical communication network robustness analysis method based on interpretation degree and high survivability algorithm aiming at the defects of the prior art. The method adopts the probability model to express the relation between the network node fault and the symptom, so that the optimal fault hypothesis set is simply and effectively judged, and the troubleshooting of the fault and the judgment on the network robustness can be more accurate; and a survivability high-survivability anti-damage algorithm is added to further judge the anti-damage capability of the network, so that the robustness of the whole network is effectively analyzed and is more reliable.

The technical scheme for realizing the purpose of the invention is as follows:

the method relates to node fault location and node survival guarantee, wherein the node fault location uses a probability weighted bipartite model, defines parameter Bayes symptom interpretation degree, represents the relationship between network node fault and symptom, calculates parameter Bayes suspicion degree as the selection standard of possible fault, and realizes the fault location algorithm based on the Bayes symptom interpretation degree; the node survival guarantee calculates the probability of abnormal communication between nodes according to the fault probability of the optical network, calculates the anti-destruction capability of the network according to the proportion of the completed service volume in the case of the network fault to the completed service volume in the normal working state, outputs the anti-destruction capability of the network, and finally synthesizes the robustness of the reflection network of the two calculation result sides;

the method specifically comprises the following steps:

1) and node fault positioning:

expressing the relationship between the link failure and the symptom according to a Probability Weighted Bipartite Graph (PWBG) model, wherein a binary node combination consisting of the link failure and the symptom is represented as V, as shown in formula (1):

V＝F∪S (1)，

wherein F is a fault set, and S is a symptom set; and defining the directed edge set of the fault pointing sign as E, and then E is F multiplied by S;

defining the probability value of the occurrence of the symptom s under the condition of the occurrence of the fault f as P (s | f), and in the deterministic model, P is_F×S(0,1) denotes determinationA set of (a); in the non-deterministic model, P_F×S(0,1) represents in the range of 0 to 1;

further study of PWBG, defining parameter F(s)_i)、S(f_i) And S_OWherein F(s)_i) Representation and symptom s_iSet of all associated faults, S (f)_i) Presentation and failure f_iThe set of all the symptoms of the association,

representing a set of symptoms observable in an optical communications network link system;

calculating a selection standard of a possible fault according to a Bayesian selected regression (BSD for short) fault location algorithm, wherein the probability of the fault is more accurately expressed by the definition of the BSD ratio form, the accuracy of the parameter is related to the symptom number corresponding to the fault, and the less the symptom number corresponding to the fault is, the coarser the parameter value is, and the larger the difference between the parameter value and the probability value of the actual fault is; the more the number of corresponding symptoms of the fault is, the more accurate the value of the parameter is, and the smaller the difference between the value of the probability and the actual fault occurrence probability is; when the number of symptoms corresponding to a plurality of possible faults related to the same symptom is large, the accuracy of possibility estimation of different faults by the parameter is large, and at the moment, fault misjudgment is easy to occur; this situation is more likely to occur in small scale networks. Therefore, Bayesian Symptom Explained Degree (BSED) is introduced, under a probability weighted bipartite graph model, the obtained Bayesian posterior probability is processed step by step, and a parameter Bayesian symptom explained degree alpha is defined, wherein the Bayesian symptom explained degree not only reflects the difference of one symptom to a plurality of related fault feedback information, but also reflects the difference of a plurality of symptoms to the same related fault providing information, more accurately expresses the probability of fault occurrence, and the specific steps of calculating the Bayesian symptom explaination degree are as follows:

the posterior probability p (f) is calculated according to the Bayes formula_j|s_i) As shown in equation (2):

wherein s is_i∈S_N，f_j∈F(s_i)，p(f_j|s_i) Indicates a symptom s_iIn the presence of a fault f_jProbability of occurrence, p (f)_j|s_i) The larger the value, the fault f_jInterpretation of symptoms s_iThe greater the probability of (a), p(s)_i|f_j) Representing the fault f as a priori probability_jUnder the conditions occurring, symptoms s_iThe probability of occurrence;

for different faults f_j∈F(s_i) Posterior probability value p (f)_j|s_i) Also different, therefore, different faults f_jFor different symptoms s_iIs different, the failure f is obtained by normalization operation_jFor symptom s_iDegree of interpretation of (a) (f)_j,s_i) Not only can guarantee the symptom s_iCan be fault set F(s)_i) May also express the selection of a certain fault f_jTo explain the symptoms s_iFrom a statistical point of view, a (f)_j,s_i) Can be interpreted as a fault f_jThe symptom s can be interpreted_iThe number of (2); the specific steps of the normalization operation are as follows:

for each symptom s_i(s_i∈S_N) The calculated posterior probability p (f)_j|s_i) Carrying out normalized calculation to obtain a fault f_jFor symptom s_iDegree of interpretation of (a) (f)_j,s_i) As shown in equation (3):

calculating the possible faults F (F is larger than F) to a symptom set S one by one according to a formula (3)_NBayesian symptom solution ofDegree of release alpha (f, S)_N)，α(f,S_N) The size of the value indicates a fault f interpreting a set of symptoms S_NThe number of medium signs, α (f, S)_N) The larger the value, the greater the probability of failure occurrence, as shown in equation (4):

finally, for set F (S)_N) The Bayesian symptom interpretation degree alpha (f, S) of each fault f_N) Form a set F_αAnd to F_αThe middle elements are sorted from large to small when F_αThe first m most likely faults in the set completely cover all observed symptom sets S_NIf so, determining that an optimal fault hypothesis set is found;

2) and (3) node survivability guarantee:

in the operation process of the optical network, whether natural disasters or artificial damages are random and uncertain, and specific damage information of the optical network is difficult to obtain. For such random faults, because the occurrence of the faults has uncertainty and there is a possibility of the faults occurring in any segment of the optical fiber link, the probability of the faults occurring is considered to be the same; meanwhile, when any section of link in the optical fiber fails, the failure of other sections cannot be influenced, and the failures are independent. Therefore, it can be seen that random faults act on the optical fiber link uniformly and independently, and it can be assumed that the probability of the random fault occurring is λ, and the length Δ L of the optical fiber link is infinitesimal, so that the probability p (Δ L) of the optical fiber link occurring fault can be calculated, as shown in equation (5):

p(ΔL)=λΔL (5)，

then, the optical fiber link with the length of L is divided into N links with the length of Δ L, and when each link does not fail, the link is in a normal working state, so that by using the limit solution principle, the probability of the normal working of the optical fiber link with the length of L can obtain the probability of failure p (L), as shown in formula (6):

considering that transmission interruption between Optical fiber transmission nodes can cause loss of a large amount of data, in order to improve network survivability, a 1+1 hot backup protection mode is adopted in a broadband Passive Optical integrated access standard (GPON) access network, when a fault occurs, an Optical Line Terminal (OLT) learns a main Optical fiber fault by detecting link information, then protection switching is automatically initiated, and standby Optical fiber transmission data is started, so that the network link protection mode is taken into consideration, when the lengths of the main and standby Optical fibers are L, the probability P that a link between the OLT and an Optical splitter normally works can be calculated by the probability that the main and standby Optical fibers simultaneously break down₀As shown in equation (7):

P₀＝1-(1-e^-λL)² (7)，

then, considering that an Optical Network can use an Optical Network Unit (Optical Network Unit, abbreviated as ONU) with a wireless function as Massive Parallel Processing (MPP) to complete data transmission between a wireless domain node and an Optical domain node, and an ONU without a wireless function is used for carrying services in a conventional Optical Network, taking a wireless transmission mode of the Optical Network into consideration, considering that the ONUs with wireless functions can establish connection through a wireless path, thereby providing protection for a branch Optical fiber connected with the ONU, however, when the branch Optical fiber connected with the ONU without a wireless function is in a fault state, data cannot be forwarded due to the absence of a redundant path, in order to know the situation that the branch Optical fiber in the Network is protected, wireless function deployment information of the ONU in the Network needs to be obtained, and a binary constant is introduced to describe the deployment situation of the ONU, whether the current ONU performs wireless communication function deployment is represented by the following formula (8):

in the optical network, the front-end wireless network can provide protection for the branch optical fiber, that is, when the branch optical fiber fails, if the ONU connected to the branch optical fiber has a wireless function, data affected by the failure can be sent to other ONUs through the wireless node at the front end, and therefore, the probability P that the ONU and the optical splitter cannot normally communicate (the ONU and the branch optical fiber having the wireless function simultaneously fail) can be obtained by calculating the probability that the branch optical fiber connected to the ONU and the branch optical fiber having the wireless function fail simultaneously_i) As shown in formula (9):

as can be seen from the formula (9), when the parameter x is_iAnd x_jWhen the sum is 1, the ONU is indicated_iAnd ONU_jAll have wireless communication function deployment, and ONU_iAnd ONU_jCan carry on the data transmission through the wireless route in between;

in the data transmission process between the ONU and the OLT, the ONU transmits the data to the optical splitter and then is forwarded to the OLT by the optical splitter, and according to the obtained probability that the OLT and the optical splitter can not normally communicate and the probability that the ONU and the optical splitter can not normally communicate, under the action of a fault, the ONU_iProbability P 'of normal communication with OLT'_iAs shown in equation (10):

because the services from the user end in the optical network are all converged into the ONU, and the uplink channel from the ONU to the OLT adopts a time division multiple access mechanism, the services from the ONU are all transmitted into the OLT; the downlink direction from the OLT to the ONUs adopts a broadcast mechanism, and the OLT service is distributed to all ONUs, so that it can be seen that the service distribution type in the back-end optical domain in the optical network belongs to a centralized type, assuming that the service from the user side is to be transmitted to the ONUs, and at this time, there is a fault in the link between the ONUs and the OLT, and there is a situation of service data loss, and the traffic that can be successfully transmitted from the ONUs to the OLT finally is the traffic that can be completed by the network, and assuming that the traffic from the ONUs are the same, and the traffic that can be completed by the network is related to the communication probability between the ONUs and the OLT, the total traffic S (λ) from the ONUs can be calculated, as shown in formula (11):

wherein, P'_iIndicating an optical network unit ONU_iProbability of communication with OLT, S_i(λ) represents an ONU from a certain optical network unit_iTraffic of (2);

the survivability of the optical network is expressed as the survivability of the network, when the proportion of the service volume finished by the network under the condition of fault is larger than that of the service volume finished under the normal working state, the survivability of the network is shown to be stronger, the production strength of the network is also higher, the survivability of the network is quantized according to the definition of the survivability, the obtained result is called as the reliability of the network, and the specific calculation method of the reliability V (lambda) of the network can be known according to the definition of the survivability is the ratio of the service volume finished under the condition of network fault to the service volume finished under the normal working state of the network, as shown in the formula (12):

the method comprises the following steps that N represents the number of ONU in an optical network, S (lambda) represents the traffic which can be completed by the network, and the network reliability is the average value of the normal communication probability between the ONU and the OLT, so that the higher the probability of normal communication between the ONU and the OLT is, the stronger the network reliability is;

finally, a set of fault hypotheses F is output_αAnd optical network reliability V (lambda), synthetic fault hypothesis set F_αAnd obtaining the result of the optical network reliability V (lambda) to obtain the robustness analysis of the optical communication network.

The method combines Bayesian symptom interpretation degree and survivability high-survivability algorithm to analyze the overall robustness of the optical communication network, can more accurately judge the fault hypothesis point, and can comprehensively judge more reliable network robustness by combining the calculated network survivability.

Drawings

FIG. 1 is a schematic diagram of a PWBG in an example embodiment;

FIG. 2 is a schematic flow chart of an embodiment.

Detailed Description

The invention is described in further detail below with reference to the following figures and specific examples, but the invention is not limited thereto.

Example (b):

referring to fig. 2, the method for analyzing robustness of an optical communication network based on interpretation degree and high survivability algorithm relates to node fault location and node survival guarantee, wherein the node fault location uses a probability weighted bipartite model, defines parameter bayesian symptom interpretation degree, represents the relationship between network node fault and symptom, calculates parameter bayesian suspicion degree as the selection standard of possible fault, and realizes the fault location algorithm based on the bayesian symptom interpretation degree; the node survival guarantee calculates the probability of abnormal communication between nodes according to the fault probability of the optical network, calculates the anti-destruction capability of the network according to the proportion of the completed service volume in the case of the network fault to the completed service volume in the normal working state, outputs the anti-destruction capability of the network, and finally synthesizes the robustness of the reflection network of the two calculation result sides;

the method specifically comprises the following steps:

1) and node fault positioning:

expressing the relation between the link failure and the symptom according to a PWBG model, wherein the combination of two nodes consisting of the link failure and the symptom is represented as V, and the formula (1) is shown as follows:

V＝F∪S (1)，

wherein F is a fault set, and S is a symptom set; and defining the directed edge set of the fault pointing sign as E, and then E is F multiplied by S;

defining the probability value of the occurrence of the symptom s under the condition of the occurrence of the fault f as P (s | f), and in the deterministic model, P is_F×S{0,1} represents a certain set; in the non-deterministic model, P_F×S(0,1) represents in the range of 0 to 1;

further study of PWBG, defining parameter F(s)_i)、S(f_i) And S_OWherein F(s)_i) Representation and symptom s_iSet of all associated faults, S (f)_i) Presentation and failure f_iThe set of all the symptoms of the association,

representing a set of symptoms observable in an optical communications network link system, a schematic diagram of a PWBG is shown in fig. 1;

calculating a selection standard of possible faults according to the BSD, wherein the probability of the faults is more accurately expressed by the definition of the ratio form of the BSD, the accuracy of the parameters is related to the number of symptoms corresponding to the faults, and the less the number of the symptoms corresponding to the faults is, the coarser the value of the parameters is, and the larger the difference between the value of the probability of the faults and the actual probability value of the faults is; the more the number of corresponding symptoms of the fault is, the more accurate the value of the parameter is, and the smaller the difference between the value of the probability and the actual fault occurrence probability is; when the number of symptoms corresponding to a plurality of possible faults related to the same symptom is large, the accuracy of possibility estimation of different faults by the parameter is large, and at the moment, fault misjudgment is easy to occur; this situation is more likely to occur in small scale networks. Therefore, BSED is introduced to carry out step processing on the obtained Bayesian posterior probability under a probability weighted bipartite graph model, and a parameter Bayesian symptom interpretation degree alpha is defined, wherein the Bayesian symptom interpretation degree not only reflects the difference of one symptom on feedback information of a plurality of related faults, but also reflects the difference of a plurality of symptoms on information provided by the same related faults, and the probability of fault occurrence is more accurately expressed, and the specific steps of calculating the Bayesian symptom interpretation degree are as follows:

the posterior probability p (f) is calculated according to the Bayes formula_j|s_i) As shown in equation (2):

wherein s is_i∈S_N，f_j∈F(s_i)，p(f_j|s_i) Indicates a symptom s_iIn the presence of a fault f_jProbability of occurrence, p (f)_j|s_i) The larger the value, the fault f_jInterpretation of symptoms s_iThe greater the probability of (a), p(s)_i|f_j) Representing the fault f as a priori probability_jUnder the conditions occurring, symptoms s_iThe probability of occurrence;

for different faults f_j∈F(s_i) Posterior probability value p (f)_j|s_i) Also different, therefore, different faults f_jFor different symptoms s_iIs different, the failure f is obtained by normalization operation_jFor symptom s_iDegree of interpretation of (a) (f)_j,s_i) Not only can guarantee the symptom s_iCan be fault set F(s)_i) May also express the selection of a certain fault f_jTo explain the symptoms s_iFrom a statistical point of view, a (f)_j,s_i) Can be interpreted as a fault f_jThe symptom s can be interpreted_iThe number of (2);

the specific steps of the normalization operation are as follows:

for each symptom s_i(s_i∈S_N) The calculated posterior probability p (f)_j|s_i) Carrying out normalized calculation to obtain a fault f_jFor symptom s_iDegree of interpretation of (a) (f)_j,s_i) As shown in equation (3):

calculating the possible faults F (F is larger than F) to a symptom set S one by one according to a formula (3)_NBayesian symptom interpretation degree alpha (f, S) of_N)，α(f,S_N) Of valueSize-indicating failure f interpretation symptom set S_NThe number of medium signs, α (f, S)_N) The larger the value, the greater the probability of failure occurrence, as shown in equation (4):

finally, for set F (S)_N) The Bayesian symptom interpretation degree alpha (f, S) of each fault f_N) Form a set F_αAnd to F_αThe middle elements are sorted from large to small when F_αThe first m most likely faults in the set completely cover all observed symptom sets S_NIf so, determining that an optimal fault hypothesis set is found;

2) and (3) node survivability guarantee:

in the operation process of the optical network, whether natural disasters or artificial damages are random and uncertain, and specific damage information of the optical network is difficult to obtain. For such random faults, because the occurrence of the faults has uncertainty and there is a possibility of the faults occurring in any segment of the optical fiber link, the probability of the faults occurring is considered to be the same; meanwhile, when any section of link in the optical fiber fails, the failure of other sections cannot be influenced, and the failures are independent. Therefore, it can be seen that random faults act on the optical fiber link uniformly and independently, and it can be assumed that the probability of the random fault occurring is λ, and the length Δ L of the optical fiber link is infinitesimal, so that the probability p (Δ L) of the optical fiber link occurring fault can be calculated, as shown in equation (5):

p(ΔL)＝λΔL (5)，

then, the optical fiber link with the length of L is divided into N links with the length of Δ L, and when each link does not fail, the link is in a normal working state, so that by using the limit solution principle, the probability of the normal working of the optical fiber link with the length of L can obtain the probability of failure p (L), as shown in formula (6):

considering that transmission interruption between optical fiber transmission nodes can cause loss of a large amount of data, in order to improve network survivability, a 1+1 hot backup protection mode is adopted in a GPON access network, when an OLT learns a main optical fiber fault by detecting link information, protection switching is automatically initiated, and backup optical fiber transmission data is started, so that the network link protection mode is taken into consideration, when the lengths of main and backup optical fibers are L, the probability of simultaneous fault of the main and backup optical fibers can be used for calculating the probability P of normal work of a link between the OLT and an optical splitter₀As shown in equation (7):

P₀＝1-(1-e^-λL)² (7)，

then considering that the optical network may use the ONU with wireless function as the MPP to complete data transmission between the wireless domain node and the optical domain node, the ONUs without wireless function are used for carrying the service in the traditional optical network, the wireless transmission mode of the optical network is taken into consideration, the ONUs with wireless function can establish connection through a wireless path, thereby providing protection for the spur optical fiber to which the ONU is connected, whereas a spur optical fiber connected to an ONU without radio functionality, in its failure state, since no redundant path exists, data cannot be forwarded, and in order to know the protection condition of the branch optical fiber in the network, wireless function deployment information of the ONU in the network is required to be obtained, a binary constant is introduced for describing the deployment situation of the ONU, whether the current ONU performs wireless communication function deployment is represented by the following formula (8):

in the optical network, the front-end wireless network can provide protection for the branch optical fiber, that is, when the branch optical fiber fails, if the ONU connected with the branch optical fiber has the wireless function, the data affected by the failure can be sent to other ONUs through the wireless node at the front end, so that the branch connected with the ONU is calculatedThe probability that the optical fiber and the branch optical fiber with the wireless function simultaneously fail can be obtained, so that the probability P (ONU) that the ONU and the optical splitter cannot normally communicate with each other can be obtained_i) As shown in formula (9):

as can be seen from the formula (9), when the parameter x is_iAnd x_jWhen the sum is 1, the ONU is indicated_iAnd ONU_jAll have wireless communication function deployment, and ONU_iAnd ONU_jCan carry on the data transmission through the wireless route in between;

in the data transmission process between the ONU and the OLT, the ONU transmits the data to the optical splitter and then is forwarded to the OLT by the optical splitter, and according to the obtained probability that the OLT and the optical splitter can not normally communicate and the probability that the ONU and the optical splitter can not normally communicate, under the action of a fault, the ONU_iProbability P 'of normal communication with OLT'_iAs shown in equation (10):

because the services from the user end in the optical network are all converged into the ONU, and the uplink channel from the ONU to the OLT adopts a time division multiple access mechanism, the services from the ONU are all transmitted into the OLT; the downlink direction from the OLT to the ONUs adopts a broadcast mechanism, and the OLT service is distributed to all ONUs, so that it can be seen that the service distribution type in the back-end optical domain in the optical network belongs to a centralized type, assuming that the service from the user side is to be transmitted to the ONUs, and at this time, there is a fault in the link between the ONUs and the OLT, and there is a situation of service data loss, and the traffic that can be successfully transmitted from the ONUs to the OLT finally is the traffic that can be completed by the network, and assuming that the traffic from the ONUs are the same, and the traffic that can be completed by the network is related to the communication probability between the ONUs and the OLT, the total traffic S (λ) from the ONUs can be calculated, as shown in formula (11):

wherein, P'_iIndicating an optical network unit ONU_iProbability of communication with OLT, S_i(λ) represents an ONU from a certain optical network unit_iTraffic of (2);

the survivability of the optical network is expressed as the survivability of the network, when the proportion of the service volume finished by the network under the condition of fault is larger than that of the service volume finished under the normal working state, the survivability of the network is shown to be stronger, the production strength of the network is also higher, the survivability of the network is quantized according to the definition of the survivability, the obtained result is called as the reliability of the network, and the specific calculation method of the reliability V (lambda) of the network can be known according to the definition of the survivability is the ratio of the service volume finished under the condition of network fault to the service volume finished under the normal working state of the network, as shown in the formula (12):

the method comprises the following steps that N represents the number of ONU in an optical network, S (lambda) represents the traffic which can be completed by the network, and the network reliability is the average value of the normal communication probability between the ONU and the OLT, so that the higher the probability of normal communication between the ONU and the OLT is, the stronger the network reliability is;

finally, a set of fault hypotheses F is output_αAnd optical network reliability V (lambda), synthetic fault hypothesis set F_αAnd obtaining the result of the optical network reliability V (lambda) to obtain the robustness analysis of the optical communication network.

Claims (1)

1. The optical communication network robustness analysis method based on the interpretation degree and the high survivability algorithm is characterized by involving node fault location and node survival guarantee, wherein the node fault location uses a probability weighted bipartite model, defines parameter Bayesian symptom interpretation degree, represents the relationship between network node faults and symptoms, calculates the parameter Bayesian suspicion degree as the selection standard of possible faults, and realizes the fault location algorithm based on the Bayesian symptom interpretation degree; the node survival guarantee calculates the probability of abnormal communication between nodes according to the fault probability of the optical network, calculates the anti-destruction capability of the network according to the proportion of the completed service volume in the case of the network fault to the completed service volume in the normal working state, outputs the anti-destruction capability of the network, and finally synthesizes the robustness of the reflection network of the two calculation result sides;

the method specifically comprises the following steps:

1) and node fault positioning:

expressing the relation between the link failure and the symptom according to a PWBG model, wherein the combination of two nodes consisting of the link failure and the symptom is represented as V, and the formula (1) is shown as follows:

V＝F∪S (1)，

wherein F is a fault set, and S is a symptom set; and defining the directed edge set of the fault pointing sign as E, and then E is F multiplied by S;

defining the probability value of the occurrence of the symptom s under the condition of the occurrence of the fault f as P (s | f), and in the deterministic model, P is_F×S{0,1} represents a certain set; in the non-deterministic model, P_F×S(0,1) represents in the range of 0 to 1;

further study of PWBG, defining parameter F(s)_i)、S(f_i) And S_OWherein F(s)_i) Representation and symptom s_iSet of all associated faults, S (f)_i) Presentation and failure f_iThe set of all the symptoms of the association,

representing a set of symptoms observable in an optical communications network link system;

calculating a selection standard of possible faults according to the BSD, wherein the number of symptoms corresponding to the faults is smaller, the parameter value is rougher, and the difference between the parameter value and the probability value of the actual faults is larger; the more the number of corresponding symptoms of the fault is, the more accurate the value of the parameter is, and the smaller the difference between the value of the probability and the actual fault occurrence probability is; when the number of symptoms corresponding to a plurality of possible faults related to the same symptom is large, the accuracy of possibility estimation of different faults by the parameter is large, and at the moment, fault misjudgment is easy to occur; therefore, the BSED is introduced to carry out step processing on the obtained Bayes posterior probability under a probability weighted bipartite graph model, a parameter Bayes symptom interpretation degree alpha is defined, and the specific steps of calculating the Bayes symptom interpretation degree are as follows:

the posterior probability p (f) is calculated according to the Bayes formula_j|s_i) As shown in equation (2):

wherein s is_i∈S_N，f_j∈F(s_i)，p(f_j|s_i) Indicates a symptom s_iIn the presence of a fault f_jProbability of occurrence, p (f)_j|s_i) The larger the value, the fault f_jInterpretation of symptoms s_iThe greater the probability of (a), p(s)_i|f_j) Representing the fault f as a priori probability_jUnder the conditions occurring, symptoms s_iThe probability of occurrence;

for different faults f_j∈F(s_i) Posterior probability value p (f)_j|s_i) Also different, therefore, different faults f_jFor different symptoms s_iIs different, the failure f is obtained by normalization operation_jFor symptom s_iDegree of interpretation of (a) (f)_j,s_i) Not only can guarantee the symptom s_iCan be fault set F(s)_i) May also express the selection of a certain fault f_jTo explain the symptoms s_iFrom a statistical point of view, a (f)_j,s_i) Can be interpreted as a fault f_jThe symptom s can be interpreted_iThe number of (2);

the normalization operation comprises the following specific steps:

for each symptom s_i(s_i∈S_N) The calculated posterior probability p (f)_j|s_i) Carrying out normalized calculation to obtain a fault f_jFor symptom s_iDegree of interpretation of (a) (f)_j,s_i) As shown in equation (3):

calculating the possible faults F (F is larger than F) to a symptom set S one by one according to a formula (3)_NBayesian symptom interpretation degree alpha (f, S) of_N)，α(f,S_N) The size of the value indicates a fault f interpreting a set of symptoms S_NThe number of medium signs, α (f, S)_N) The larger the value, the greater the probability of failure occurrence, as shown in equation (4):

finally, for set F (S)_N) The Bayesian symptom interpretation degree alpha (f, S) of each fault f_N) Form a set F_αAnd to F_αThe middle elements are sorted from large to small when F_αThe first m most likely faults in the set completely cover all observed symptom sets S_NIf so, determining that an optimal fault hypothesis set is found;

2) and (3) node survivability guarantee:

assuming that the probability of occurrence of random failure is λ, and the length Δ L of the optical fiber link is infinitesimally small, the probability p (Δ L) of occurrence of failure of the optical fiber link can be calculated, as shown in equation (5):

p(ΔL)＝λΔL (5)，

then, the optical fiber link with the length of L is divided into N links with the length of Δ L, and when each link does not fail, the link is in a normal working state, so that by using the limit solution principle, the probability of the normal working of the optical fiber link with the length of L can obtain the probability of failure p (L), as shown in formula (6):

considering that transmission interruption between optical fiber transmission nodes can cause loss of a large amount of data, in order to improve network survivability, a 1+1 hot backup protection mode is adopted in a GPON access network, when an OLT learns a main optical fiber fault by detecting link information, protection switching is automatically initiated, and backup optical fiber transmission data is started, so that the network link protection mode is taken into consideration, when the lengths of main and backup optical fibers are L, the probability of simultaneous fault of the main and backup optical fibers can be used for calculating the probability P of normal work of a link between the OLT and an optical splitter₀As shown in equation (7):

P₀＝1-(1-e^-λL)² (7)，

then considering that the optical network may use the ONU with wireless function as the MPP to complete data transmission between the wireless domain node and the optical domain node, the ONUs without wireless function are used for carrying the service in the traditional optical network, the wireless transmission mode of the optical network is taken into consideration, the ONUs with wireless function can establish connection through a wireless path, thereby providing protection for the spur optical fiber to which the ONU is connected, whereas a spur optical fiber connected to an ONU without radio functionality, in its failure state, since no redundant path exists, data cannot be forwarded, and in order to know the protection condition of the branch optical fiber in the network, wireless function deployment information of the ONU in the network is required to be obtained, a binary constant is introduced for describing the deployment situation of the ONU, whether the current ONU performs wireless communication function deployment is represented by the following formula (8):

in the optical network, the front-end wireless network can provide protection for the branch optical fiber, namely when the branch optical fiber appearsWhen the fault occurs, if the ONU connected with the branch optical fiber has the wireless function, the data affected by the fault can be sent to other ONUs through the wireless node at the front end, so that the probability P that the ONU cannot normally communicate with the optical splitter (the ONU and the ONU) can be obtained by calculating the probability that the branch optical fiber connected with the ONU and the branch optical fiber with the wireless function simultaneously fail_i) As shown in formula (9):

as can be seen from the formula (9), when the parameter x is_iAnd x_jWhen the sum is 1, the ONU is indicated_iAnd ONU_jAll have wireless communication function deployment, and ONU_iAnd ONU_jCan carry on the data transmission through the wireless route in between;

in the data transmission process between the ONU and the OLT, the ONU transmits the data to the optical splitter and then is forwarded to the OLT by the optical splitter, and according to the obtained probability that the OLT and the optical splitter can not normally communicate and the probability that the ONU and the optical splitter can not normally communicate, under the action of a fault, the ONU_iProbability P of normal communication with OLT_i', as shown in equation (10):

because the services from the user end in the optical network are all converged into the ONU, and the uplink channel from the ONU to the OLT adopts a time division multiple access mechanism, the services from the ONU are all transmitted into the OLT; the downlink direction from the OLT to the ONUs adopts a broadcast mechanism, the OLT service is distributed to all ONUs, assuming that the service from the user side is transmitted to the ONUs, and at this time, a link between the ONUs and the OLT has a fault and a condition of service data loss exists, and the traffic that can be finally successfully transmitted from the ONUs to the OLT is the traffic that can be completed by the network, assuming that the traffic from the ONUs are the same and the traffic that can be completed by the network is related to the communication probability between the ONUs and the OLT, the total traffic S (λ) from the ONUs can be calculated, as shown in formula (11):

wherein, P_i' indicates a certain optical network Unit ONU_iProbability of communication with OLT, S_i(λ) represents an ONU from a certain optical network unit_iTraffic of (2);

the survivability of the optical network is expressed as the survivability of the network, when the proportion of the service volume finished by the network under the condition of fault is larger than that of the service volume finished under the normal working state, the survivability of the network is shown to be stronger, the production strength of the network is also higher, the survivability of the network is quantized according to the definition of the survivability, the obtained result is called as the reliability of the network, and the specific calculation method of the reliability V (lambda) of the network can be known according to the definition of the survivability is the ratio of the service volume finished under the condition of network fault to the service volume finished under the normal working state of the network, as shown in the formula (12):

the method comprises the following steps that N represents the number of ONU in an optical network, S (lambda) represents the traffic which can be completed by the network, and the network reliability is the average value of the normal communication probability between the ONU and the OLT, so that the higher the probability of normal communication between the ONU and the OLT is, the stronger the network reliability is;

finally, a set of fault hypotheses F is output_αAnd optical network reliability V (lambda), synthetic fault hypothesis set F_αAnd obtaining the result of the optical network reliability V (lambda) to obtain the robustness analysis of the optical communication network.

Images