CN110309612B - Power system fault processing method based on fuzzy fault Petri network - Google Patents

Abstract

The invention discloses a power system fault processing method based on a fuzzy fault Petri network, which comprises the following steps: establishing an incidence relation between a component and a system fault based on a fault type of a power system, establishing a fuzzy fault Petri network and a modeling rule, wherein the fuzzy fault Petri network continuously changes the state of the system according to a transition enabling rule and an occurrence rule, describes the dynamic characteristic of the system, and obtains a fault event confidence coefficient, transition ignition and an event fault inference process based on a forward inference algorithm so as to obtain a basic event confidence coefficient, a transition ignition sequence and a fault propagation path; obtaining a search path of the fault based on a reverse reasoning algorithm to obtain a rapid diagnosis strategy of the fault; and (4) realizing the processing of system-level faults based on a forward and backward reasoning diagnosis algorithm.

Description

Power system fault processing method based on fuzzy fault Petri network

Technical Field

The invention belongs to the field of fault detection, and particularly relates to a power system fault processing method based on a fuzzy fault Petri network.

Background

Armored vehicles are major army warfare weaponry for ground assault and counterassault, and the reliability of tasks is an important index of the operational performance of armored vehicles. The actual use environment of the device is complex, and the device is often in a severe environment with high speed, large load and strong vibration impact, so that the failure frequency is caused, and the maintenance cost reaches more than 72 percent of the total life cost of the device according to statistics. The power system is a multi-fault component, and the fault frequency of the power system accounts for about 51 percent of the total fault number of the whole vehicle. Due to the fact that the structure is complex, the load is variable, and the information uncertainty is strong, the failure prediction and the quick repair difficulty are increased. With the wide application of high and new technologies, the tactical performance of modern armor equipment is continuously improved, the system composition structure is increasingly complex, and the use and guarantee cost is increasingly huge. Therefore, the fault prediction and diagnosis of the armored vehicle is important work for researching the reliability, maintainability and supportability of equipment, and the accurate fault prediction and diagnosis has important theoretical and practical significance for guiding the maintenance and the guarantee of the armored vehicle, avoiding unnecessary harm caused by blind disassembly, shortening the fault identification time and the like. The power system of the armored vehicle plays an important role in playing the maneuvering performance of the vehicle, and the power system is a complex system mainly composed of 11 subsystems such as a crank link mechanism, a linkage mechanism, a gas distribution mechanism, a turbocharger, a cooling system, a lubricating system, an air supply system, a fuel supply system, a heating system, an exhaust system and a starting system. Due to the fact that the system is complex in structure, severe in use environment, variable in load and strong in information uncertainty, the failure generation mechanism of the power system is more complex, the failure rate of the power system is high, and the failure prediction and rapid repair difficulty is increased. Most of the traditional fault monitoring and diagnosis aims at a certain device or subsystem, the possible faults are deduced through the correlation analysis of a large number of alarms, the fault diagnosis only carried out on a certain device or a certain link cannot realize the high reliability and the high maintainability of the whole system from the system level, the normal operation of the system can be influenced, and the occurrence of catastrophic accidents is caused.

The above information disclosed in this background section is only for enhancement of understanding of the background of the invention and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art.

Disclosure of Invention

The invention aims to realize the purpose through the following technical scheme, and the power system fault processing method based on the fuzzy fault Petri network comprises the following steps:

in the first step, the incidence relation between the components and the system faults is established based on the fault type of the power system,

in the second step, a fuzzy fault Petri net and a modeling rule are established, the fuzzy fault Petri net continuously changes the state of the system according to the transition enabling rule and the occurrence rule and describes the dynamic characteristic of the system, wherein the fuzzy fault Petri net is defined as a 12-tuple (P, T, I, O, K, T)_t，M，w，f，α，λ，U_μ) Wherein: p ═ P₁，p₂，…，p_mIs a collection of non-empty finite libraries representing a collection of system failure events, T ═ T₁，t₂，…t_nThe method is characterized in that the method is a non-empty finite transition set, represents state change or behavior action of a simulated system, and reflects propagation evolution of internal faults of the system, and I: p × T is an input matrix, representing P_i→t_jIn the middle ofDirected arc of (i.e. transition t)_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; o: t × P is the output matrix, representing the output from T_j→p_iDirected arcs existing between, i.e. transitions t_jIs output arc of, and p_iTo transition t_jI is 1, 2, …, m; j is 1, 2, …, n; k: p → [0, 1]Is the capacity function of the fault Petri net; t is_t: triggered transition vector with initial value of T⁰ _t＝(0，0，…，0)^TThe element is set to 1 after the same fault occurs, which means that the element cannot occur again before being repaired; m ═ M₁，m₂，…，m_m)^TDistribution vectors, m, identified for the library_iIdentify its corresponding depot p_iState of (1), m₀Is the initial vector identification of the library, and represents the initial state of the simulated system; w ═ w (w)₁，w₂，…，w_n)^TIs a library site fault event weight vector, reflects the input library site p_kDegree of influence on transition t, wherein

p_kIs e.g. I and

f＝(f₁，f₂，…，f_m)^Tis a fuzzy probability threshold vector of the fault event of the library, and the confidence coefficient alpha of the fault event_i＞f_iA fault event may be considered true, otherwise it is false; α ═ α (α)₁，α₂，…，α_m)^TFor fault event confidence vectors, including α_default’α_INAnd alpha_OUT’α_defaultObtaining a default fault event confidence coefficient vector according to fault statistics, wherein the default fault event confidence coefficient vector comprises all the libraries; alpha is alpha_INFor inputting a confidence coefficient vector of a fault event, performing component fault diagnosis according to sensor monitoring data and acquiring the component fault after the component fault diagnosis, wherein the element is an initial library；α_OUTObtaining a confidence coefficient vector of the output fault event according to a forward reasoning algorithm, wherein elements are all the libraries except the initial library; λ ═ λ₁，λ₂，…，λ_n)^TA confidence threshold representing that a lower-level fault causes a higher-level fault; u shape_μ＝diag(μ₁，μ₂，…，μ_n) Denotes a transition confidence matrix, μ_jRepresents a transition t_jReliability of, mu_j∈[0，1]1, 2, …, N, defining a vector N_μ＝(μ₁，μ₂，…，μ_n)^TRepresenting a transition confidence vector;

in the third step, a fault event confidence coefficient, transition ignition and an event fault reasoning process are obtained based on a forward reasoning algorithm so as to obtain a basic event confidence coefficient, a transition ignition sequence and a fault propagation path; obtaining a search path of the fault based on a reverse reasoning algorithm to obtain a rapid diagnosis strategy of the fault; and diagnosing the system-level fault based on a forward and backward reasoning diagnosis algorithm.

In the method, in the first step, the power system is an armored vehicle power plant which comprises the following components: the system comprises a crank connecting rod mechanism, a linkage mechanism, a gas distribution mechanism, a turbocharger, a cooling system, a lubricating system, an air supply system, a fuel supply system, a heating system, an exhaust system and a starting system, wherein the correlation relation is obtained according to the hierarchy and the relativity, the hierarchy is the longitudinal direction of faults, high-level faults can be caused by low-level faults, and the low-level faults can certainly cause the high-level faults; the correlation is the "transversal" of the fault, and is a system composed of a plurality of interconnected subsystems, and the fault of the subsystem is caused by the fault propagation of the subsystem related to the fault or the next-stage subsystem.

In the method, in the second step, when p is₁Then p₂Occurrence of t^·＝{p₂Denotes if library p₁Reach a repository p only through a unique transition t₂And the last set of transitions t is only { p }₂The transition enables to satisfy α₁W μ > λ, i.e. librariesTo p is₁The actual confidence of the rule is greater than the threshold required for transition triggering, and the confidence of the fault event after triggering is alpha₂＝α₁wμ。

In the method, the second step, when p is₁And p₂… and p_nGeneration, p_kOccurrence of t^·＝{p_k}, library site p₁The actual credibility of the rule is larger than the threshold value required by the transition trigger, and the transition enable meets min (alpha)₁w₁μ，α₂w₂μ，…，α_nw_nMu) > lambda, i.e. library p₁，p₂，…，p_nThe minimum of the actual confidence levels for the rules is greater than the threshold required for transition triggering, and the confidence level alpha of the fault event after triggering is triggered_k＝min(α₁w₁μ，α₂w₂μ，…α_nw_nμ)。

In the method, the second step, when p is_jP after occurrence₁And p₂… and p_nOccurrence of t^·＝{p₁，p₂，…，p_nDenotes if library p₁，p₂，…，p_nAll can reach the library site p simultaneously through the transition t₁，p₂，…，p_nAnd the last set of transitions t is only { p }₁，p₂，…，p_nThe transition enables to satisfy α_jW μ > λ is triggered, i.e. library p₁，p₂，…，p_nThe minimum of the actual confidence levels for the rules is greater than the threshold required for transition triggering, and the confidence level alpha of the fault event after triggering is triggered₁＝α₂＝…＝α_n＝α_j*w*μ。

In the method, the second step, when p is₁And p₂… and p_nP after occurrence_kThe occurrence of this is that,

i.e. depot p₁，p₂，…，p_nActual confidence in rulesThe minimum is larger than the threshold required for triggering the transition, if the transition can satisfy max (alpha)₁w₁μ₁，α₂w₂μ₂，…，α_nw_nμ_n) λ, i.e. library p₁，p₂，…，p_nThe maximum actual reliability of the rule is triggered as long as the maximum actual reliability is larger than the threshold required by the transition trigger, and the confidence alpha of the fault event after the trigger_k＝max(α₁w₁μ₁，α₂w₂μ₂，…，α_nw_nμ_n)。

In the method, the second step, when p is_j1And p_j2… and p_jmP after occurrence_k1And p_k2… and p_knOccurrence of t^·＝{p_k1，p_k2，…，p_knDenotes if library p_j1，p_j2，…，p_jnAll can reach the library site p through transition t_k1，p_k2，…，p_knAnd the last set of transitions t is only { p }_k1，p_k2，…，p_knIf the transition is enabled to satisfy min (alpha)_j1w₁μ，α_j2w₂μ，…，α_jmw_mMu) > lambda, i.e. library p₁，p₂，…，p_nThe minimum actual confidence of the rules is larger than the threshold value required by transition triggering, and the confidence alpha of the fault event after triggering is triggered_k1＝α_k2＝…＝α_kn＝min(α_j1w₁μ，α_j2w₂μ，…α_jmw_mμ)。

In the third step, the method adopts MYCIN confidence matrix reasoning method to gradually deduce and obtain all state quantities of system components, and defines 3 operators: operator for taking large value

A, B and C are all m × n matrix, then C_ij＝max(α_ij，b_ij) I 1, 2, …, m, j 1, 2, …, n, multiplicationOperator

A, B and C are the matrix of m × q, q × n and m × n respectively, then

Direct multiplier operator: c is A, B, C is m-dimensional vector, then C_i＝a_i·b_i，i＝1，2，…，m

The confidence inference formula for an event is:

in the formula, alpha^k+1And alpha^kThe confidence degrees of the event at the k-th inference and the k + 1-th inference are respectively; o is an output matrix representing the output from t_j→p_iDirected arcs existing between, i.e. transitions t_jIs output arc of, and p_iTo transition t_jI is 1, 2, …, m; j is 1, 2, …, n; u shape_μRepresenting a transition confidence matrix, element mu_jRepresents a transition t_jReliability of, mu_j∈[0，1]，j＝1，2，…，n；I^TFor input matrix, representing from p_i→t_jDirected arcs existing between, i.e. transitions t_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; w is the weight vector of the failure event in the library, which reflects the p input into the library_kDegree of influence on transition t, wherein

p_kIs e.g. I and

is an m-dimensional vector, describes the confidence that the rule is true,

the reasoning process is as follows:

the first step is as follows: initializing parameters;

the second step is that: k is 0;

the third step: from alpha^kCalculating to obtain alpha^k+1；

The fourth step: if α is^k+1≠α^kRepeating the third step; if α is^k+1＝α^kAnd the reasoning is finished.

In the method, in the third step, the reasoning process of transition ignition is as follows:

the first step is as follows: initializing parameters;

the second step is that: according to

Calculating the transition support, wherein D^k+1，D^kRespectively identifying the potential enabling transition support degree of the event at the kth time and the kth +1 time; n is a radical of_μ＝(μ₁，μ₂，…，μ_n)^TRepresenting a transition confidence vector; alpha is alpha^kThe confidence of the k-th inference for the event; i is^TFor input matrix, representing from p_i→t_jDirected arcs existing between, i.e. transitions t_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; w is the weight vector of the failure event in the library, which reflects the p input into the library_kDegree of influence on transition t, wherein

p_kIs e.g. I and

is an m-dimensional vector describing the confidence that the rule is true;

the third step: according to

Calculating the triggered transition vector T_tIn the formula, T^k+1 _t，T^k _tRespectively identifying triggered transition vectors at the kth time and the kth +1 time; d^kIdentifying a support for the event at the kth potential enabling transition; λ ═ λ₁，λ₂，…，λ_n)^TA confidence threshold representing that a lower-level fault causes a higher-level fault; alpha is alpha^kThe confidence of the k-th inference for the event; i is^TFor input matrix, representing from p_i→t_jDirected arcs existing between, i.e. transitions t_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; f. of^kIs a fuzzy probability threshold vector of the fault event of the kth inference bank;

the fourth step: according to

Calculating a fault event vector M, where M^k+1，M^kFault event vectors for the kth and the (k + 1) th times, respectively; t is^k _tIdentifying the triggered transition vector for the kth time; o is an output matrix representing the output from t_j→p_iDirected arcs existing between, i.e. transitions t_jIs output arc of, and p_iTo transition t_jI is 1, 2, …, m; j is 1, 2, …, n;

the fifth step: if M is^k+1≠M^kRepeating the second step; if M is^k+1＝M^kAnd the reasoning is finished.

In said method, t_aIs a transition, p_i，p_j，p_kThree libraries if p_i∈^·t_a，p_j∈t_a ^·Then p is_iIs referred to as p_jTo go back to the associated repository, p_jIs referred to as p_iIf p is an immediate look-ahead associative base of_iIs p_jTo go back to the associated repository, p_jIs p_kThe immediate backtracking of the associated library is called p_iIs p_kIf p is a backtracking relationship of_i∈^·t_aAnd p is_k∈^·t_aThen is called p_iAnd p_kTo transition t_aThe adjacent library of (1) contains p_iThe set of immediate backtracking associative banks is called p_iIs collected as IHS (p)_i) (ii) a Comprising p_iThe set of backtracking associative libraries is called p_iIs collected as HS (p)_i) And the reverse reasoning process:

the first step is as follows: initializing parameters;

the second step is that: finding a terminating node p_jIt is used as the initial target repository (p)_j，IHS(p_j) Whereinsaid: p is a radical of_jIs the place of the target library,^·p_j＝t_j，IHS(p_j) Is the target depot p_jTo compute an initial target library IHS (p)_j) Failure incidence of all elements;

the third step: searching a library p corresponding to the maximum fault susceptibility in all unsearched paths_kIt is used as the initial target repository (p)_k，IHS(p_k) If IHS (p)_k) If the node is an empty set, the node is a termination node, the node returns to the immediate look-ahead association library, the third step is repeated, and otherwise the next step is carried out;

the fourth step: if IHS (p)_k) And if not, proceeding downwards according to the method of the third step until all the libraries are traversed.

Compared with the prior art, the invention has the following advantages:

the method is simple and easy to implement, breaks through the nondeterministic fault propagation modeling technology of complex equipment, and solves the difficult problems of characterization and description of a fault model.

Drawings

Various other advantages and benefits of the present invention will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. It is obvious that the drawings described below are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort. Also, like parts are designated by like reference numerals throughout the drawings.

In the drawings:

FIG. 1 is a schematic diagram of the steps of the power system fault processing method based on the fuzzy fault Petri net;

FIG. 2 is a schematic structural hierarchy diagram of a power device of an armored vehicle based on the fuzzy fault Petri net power system fault processing method;

3(a) -3 (b) are schematic diagrams of the type of 'one-in-one-result' of the fuzzy fault Petri net based power system fault handling method of the invention;

4(a) -4 (b) are schematic diagrams of the type of "multi-factor one-fruit" of the fuzzy fault Petri net based power system fault handling method of the invention;

5(a) -5 (b) are schematic diagrams of the type of "one-cause-many-effect" of the fuzzy fault Petri net based power system fault handling method of the invention;

6(a) -6 (b) are schematic diagrams of the type of "competition mode" of the fuzzy fault Petri net based power system fault handling method of the invention;

7(a) -7 (b) are schematic diagrams of the type of "multi-factor and multi-effect" of the fuzzy fault Petri net based power system fault handling method of the invention;

FIG. 8 is a schematic diagram of an event confidence reasoning process of the fuzzy fault Petri net based power system fault processing method;

FIG. 9 is a schematic diagram of an intelligent diagnosis reasoning process of the fuzzy fault Petri net based power system fault processing method;

FIG. 10 is a schematic diagram of a fuzzy fault Petri network for engine oil temperature overhigh in the power system fault processing method based on the fuzzy fault Petri network.

The invention is further explained below with reference to the figures and examples.

Detailed Description

Specific embodiments of the present invention will be described in more detail below with reference to the accompanying drawings, fig. 1-10. While specific embodiments of the invention are shown in the drawings, it should be understood that the invention may be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

It should be noted that certain terms are used throughout the description and claims to refer to particular components. As one skilled in the art will appreciate, various names may be used to refer to a component. This specification and claims do not intend to distinguish between components that differ in name but not function. In the following description and in the claims, the terms "include" and "comprise" are used in an open-ended fashion, and thus should be interpreted to mean "include, but not limited to. The description which follows is a preferred embodiment of the invention, but is made for the purpose of illustrating the general principles of the invention and not for the purpose of limiting the scope of the invention. The scope of the present invention is defined by the appended claims.

For the purpose of facilitating understanding of the embodiments of the present invention, the following description will be made by taking specific embodiments as examples with reference to the accompanying drawings, and the drawings are not to be construed as limiting the embodiments of the present invention.

For better understanding, fig. 1 is a schematic flow chart of a power system fault handling method based on a fuzzy fault Petri net according to an embodiment of the invention, and as shown in fig. 1, the power system fault handling method based on the fuzzy fault Petri net comprises the following steps:

in a first step S100, an association relationship between a component and a system fault is established based on the fault type of the power system,

in a second step S200, a fuzzy fault Petri net and a modeling rule are established, the fuzzy fault Petri net continuously changes the state of the system according to a transition enabling rule and a generating rule, and describes the dynamic characteristics of the system, wherein,

the fuzzy fault Petri net is defined as 12 tuples (P, T, I, O, K, T)_t，M，w，f，α，λ，U_μ) Wherein: p ═ P₁，p₂，…，p_mIs a collection of non-empty finite libraries representing a collection of system failure events, T ═ T₁，t₂，…t_nThe method is characterized in that the method is a non-empty finite transition set, represents state change or behavior action of a simulated system, and reflects propagation evolution of internal faults of the system, and I: p × T is an input matrix, representing P_i→t_jDirected arcs existing between, i.e. transitions t_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; o: t × P is the output matrix, representing the output from T_j→p_iDirected arcs existing between, i.e. transitions t_jIs output arc of, and p_iTo transition t_jI is 1, 2, …, m; j is 1, 2, …, n; k: p → [0, 1]Is the capacity function of the fault Petri net; t is_t: triggered transition vector with initial value of T⁰ _t＝(0，0，…，0)^TThe element is set to 1 after the same fault occurs, which means that the element cannot occur again before being repaired; m ═ M₁，m₂，…，m_m)^TDistribution vectors, m, identified for the library_iIdentify its corresponding depot p_iState of (1), m₀Is the initial vector identification of the library, and represents the initial state of the simulated system; w ═ w (w)₁，w₂，…，w_n)^TIs a library site fault event weight vector, reflects the input library site p_kDegree of influence on transition t, wherein

p_kIs e.g. I and

f＝(f₁，f₂，…，f_m)^Tis a fuzzy probability threshold vector of the fault event of the library, and the confidence coefficient alpha of the fault event_i＞f_iA fault event may be considered true, otherwise it is false; α ═ α (α)₁，α₂，…，α_m)^TFor fault event confidence vectors, including α_default，α_INAnd alpha_OUT，α_defaultObtaining a default fault event confidence coefficient vector according to fault statistics, wherein the default fault event confidence coefficient vector comprises all the libraries; alpha is alpha_INObtaining a component fault diagnosis according to sensor monitoring data after inputting a fault event confidence coefficient vector, wherein the element is an initial library; alpha is alpha_OUTObtaining a confidence coefficient vector of the output fault event according to a forward reasoning algorithm, wherein elements are all the libraries except the initial library; λ ═ λ₁，λ₂，…，λ_n)^TA confidence threshold representing that a lower-level fault causes a higher-level fault; u shape_μ＝diag(μ₁，μ₂，…，μ_n) Denotes a transition confidence matrix, μ_jRepresents a transition t_jReliability of, mu_j∈[0，1]J 1, 2, …, N, defining a vector N_μ＝(μ₁，μ₂，…，μ_n)^TRepresenting a transition confidence vector;

in the third step S300, a fault event confidence, a transition firing sequence and a fault propagation path are obtained based on a forward reasoning algorithm to obtain a basic event confidence, a transition firing sequence and a fault propagation path; obtaining a search path of the fault based on a reverse reasoning algorithm to obtain a rapid diagnosis strategy of the fault; and diagnosing the system-level fault based on a forward and backward reasoning diagnosis algorithm.

The invention is particularly suitable for fault treatment of a power device of an armored vehicle, and in the embodiment, the fault treatment method of the power system based on the fuzzy fault Petri network comprises the following steps:

1) analyzing different fault types of the armored vehicle power device, determining the incidence relation between key components and system faults, describing the relation among systems, subsystems and equipment from a system level, and analyzing the fault mode and the fault level of the armored vehicle power device by considering the relation among the whole system, parts, structures and levels.

2) Defining a fuzzy fault Petri net and determining a modeling rule.

3) A forward reasoning algorithm is proposed to obtain a fault event confidence coefficient, a transition ignition and a reasoning process of event faults, and a basic event confidence coefficient, a transition ignition sequence and a fault propagation path are obtained; a reverse reasoning algorithm is proposed to obtain a search path of the fault and obtain a rapid diagnosis strategy of the fault; and a forward and backward reasoning intelligent diagnosis algorithm is provided to realize the diagnosis of system-level faults.

The step 1) comprises the following specific steps:

firstly, analyzing the structure of the power device of the armored vehicle to obtain 11 subsystem structures which comprise a crank link mechanism, a linkage mechanism, a gas distribution mechanism, a turbocharger, a cooling system, a lubricating system, an air supply system, a fuel supply system, a heating system, an exhaust system, a starting system and the like and are mutually influenced, and then analyzing the hierarchy and the relevance of the faults of the power system of the armored vehicle. Hierarchical, i.e., "longitudinal" to failure, a high level failure may be caused by a low level failure that must cause a high level failure; the correlation, namely the 'transversal' of the fault, is a system composed of a plurality of interconnected subsystems, and the fault of some subsystems is often caused by the fault propagation of the related subsystem or the next-level subsystem, thereby showing the correlation. Finally, the fault level of the armored vehicle power device is obtained, and the figure 2 is shown. The failure modes of the power device are counted and analyzed, and common failure modes are obtained and are shown in the table 1.

TABLE 1 common faults of power system of armored car

The failure modes from the complete machine to each subsystem to each component unit node are shown in table 2.

TABLE 2 failure modes of various subsystems of the power system of an armored vehicle

(a) Crank connecting rod mechanism

(b) Transmission mechanism

Serial number	Numbering	Failure mode	Hazard classification
				1.	f201	Broken tooth of transmission gear	II
2.	f202	Abnormal wear of tooth surface of transmission gear	II
				3.	f203	Fatigue crack of transmission case	III
4.	f204	Bearing wear	III
				5.	f205	Bending or breaking of shafts	II

(c) Valve train

Serial number	Numbering	Failure mode	Hazard classification
				1.	f301	Peach-shaped abrasion of distribution camshaft	II
2.	f302	Abnormal wear of rocker roller	II
				3.	f303	Breaking of rocker arm	II
4.	f304	Valve spring damage	II
				5.	f305	Valve seat ring wear	II
6.	f306	Valve wear	II
				7.	f307	Valve damage	II

(d) Supercharging intake and exhaust system

Serial number	Numbering	Failure mode	Hazard classification
				1.	f401	Compressor wheel blade fracture or damage	I
2.	f402	Loosening of back plate of compressor	II
				3.	f403	Turbine wheel breakage or damage	I
4.	f404	Loosening of compressor impeller	II
				5.	f405	Abnormal wear of rotor shaft system parts	III
6.	f406	Intermediate leakage engine oil	III
				7.	f407	Cracking and air leakage of exhaust pipe	III
8.	f408	Leakage of the exhaust gasket	III
				9.	f409	Bellows rupture	III
10.	f410	Clamp fracture	III

(e) Cooling system

Serial number	Numbering	Failure mode	Hazard classification
				1.	f501	Damage of water pump impeller	II
2.	f502	Water leakage of water pump	III
				3.	f503	The impeller of the water pump can not rotate	III
4.	f504	Cavitation of water pump	III
				5.	f505	Intercooling efficiency drop	III
6.	f506	Intercooler reveals	III

(f) Lubrication system

Serial number	Numbering	Failure mode	Hazard classification
				1.	f601	Pump cavity or gear damage of engine oil pump	II
2.	f602	Engine oil heat exchanger	III
				3.	f603	Reduction in heat dissipation efficiency of engine oil heat exchanger	III
4.	f604	Clogging or breakdown of oil filter element	III

(g) Electric control system

Serial number	Numbering	Failure mode	Hazards, etcStage
				1.	f701	Sensor failure	III
2.	f702	Pedal potentiometer failure	III
				3.	f703	ECU failure, failure of speed regulation	II
4.	f704	Virtual, open or short-circuit of cables	III
				5.	f705	Oil pressure sensor failure	III
6.	f706	Water temperature sensor failure	III
				7.	f707	Exhaust temperature sensor failure	III

(h) Fuel supply system

Serial number	Numbering	Failure mode	Hazard classification
				1.	f801	Low pressure pump failure	III
2.	f802	High pressure pump failure	III
				3.	f803	High pressure pump temperature overshoot	III
4.	f804	Untight seal of fuel filter	III
				5.	f805	Fuel filter element blockage	III
6.	f806	Fuel injector not being able to open	III
				7.	f807	Fuel injector failing to close	II
8.	f808	Excessive oil return of oil injector	III
				9.	f809	The pressure regulating valve is normally open or blocked	III
10.	f810	Leakage at sealing joint of high-pressure oil pipe	III
				11.	f811	Fracture of high pressure oil pipe	III

The step 2) is specifically as follows:

firstly, defining the fuzzy fault Petri net. Aiming at the characteristics of multilevel, multi-relevance, uncertainty and the like of the armored vehicle power system fault, the fuzzy fault Petri network definition is provided by taking a fuzzy theory and a Petri network theory as cores.

The fuzzy fault Petri net (FFPetri net) is defined as a 12-tuple (P, T, I, O, K, T)_t，M，w，f，α，λ，U_μ)：

Wherein: (1) p ═ P₁，p₂，…，p_mThe set of non-empty finite libraries represents the set of system failure events.

(2)T＝{t₁，t₂，…t_nAnd the finite transition set is a non-empty finite transition set, represents the state change or behavior action of the simulated system, and reflects the propagation evolution of the internal fault of the system.

(3) I: p × T is an input matrix, representing P_i→t_jDirected arcs existing between, i.e. transitions t_jIs input arc of, and p_iTo transition t_jThe input library of (1). i is 1, 2, …, m; j is 1, 2, …, n.

(4) O: t × P is the output matrix, representing the output from T_j→p_iDirected arcs existing between, i.e. transitions t_jIs output arc of, and p_iTo transition t_jThe output depot. i is 1, 2, …, m; j is 1, 2, …, n.

(5) K: p → [0, 1] is the capacity content of the faulty Petri net.

(6)T_t: triggered transition vector with initial value of T⁰ _t＝(0，0，…，0)^TAnd the propagation path is used for identifying the fault and preventing the transition from occurring repeatedly, namely after the same fault occurs, the element is set to be 1, which means that the element cannot occur again before being repaired.

(7)M＝(m₁，m₂，…，m_m)^TDistribution vectors, m, identified for the library_iIdentify its corresponding depot p_iThe state of (1). m is₀Is the initial vector identification of the library, representing the initial state of the system being simulated.

(8)w＝(w₁，w₂，…，w_n)^TIs a library site fault event weight vector, reflects the input library site p_kDegree of influence on transition t, wherein

p_kIs e.g. I and

(9)f＝(f₁，f₂，…，f_m)^Tis a fuzzy probability threshold vector of the fault event of the library, and the confidence coefficient alpha of the fault event_i＞f_iA fault event may be considered true, otherwise it is false.

(10)α＝(α₁，α₂，…，α_m)^TIs a fault event confidence vector. Comprising alpha_default，α_INAnd alpha_OUT。α_defaultObtaining a default fault event confidence coefficient vector according to fault statistics, wherein the default fault event confidence coefficient vector comprises all the libraries; alpha is alpha_INObtaining a component fault diagnosis according to sensor monitoring data after inputting a fault event confidence coefficient vector, wherein the element is an initial library; alpha is alpha_OUTAnd acquiring elements of all the libraries except the initial library according to a forward reasoning algorithm for outputting the confidence coefficient vector of the fault event.

(11)λ＝(λ₁，λ₂，…，λ_n)^TA confidence threshold representing that a lower level fault causes a higher level fault.

(12)U_μ＝diag(μ₁，μ₂，…，μ_n) Denotes a transition confidence matrix, μ_jRepresents a transition t_jReliability of, mu_j∈[0，1]J is 1, 2, …, n. Definition vector N_μ＝(μ₁，μ₂，…，μ_n)^TA transition confidence vector is represented.

And the fuzzy fault Petri network continuously changes the state of the system according to the transition enabling rule and the occurrence rule and describes the dynamic characteristic of the system.

Definition 1: transition enable (enabled). At the current markUnder M recognition, transition t_jIs enabled if and only if:

equation (1) indicates that the actual confidence level of the rule for the library p should be greater than or equal to the threshold required for transition triggering. Equation (2) requires that the transition does not belong to the triggered set of transitions, i.e., transition t_jHas not yet been triggered.

Definition 2: the consequence of the transition is. In the Petri network theory, the triggering of the transition only changes the state of a library associated with the transition, and in the fault propagation process, the fault does not eliminate the fault reason. Therefore, the occurrence consequence of the transition of the fuzzy fault Petri net only changes the post-repository of the transition, and T is used for changing the post-repository of the transition after the transition occurs_t＝T_t+{t_j}。

Definition 3: failure is easy to occur. The product of the confidence of a fault event and the fault rate of the event is referred to as the fault susceptibility f (p)_i)。

And finally establishing a modeling rule according to the fault propagation mode. Because there are one-cause, one-cause multiple-effect, competition mode and multi-cause multiple-effect fault propagation modes in the fault propagation process, the corresponding transition occurrence result can be related to the synthetic fuzzy generation rule, that is, the precondition or conclusion part in the fuzzy generation rule contains the conjunction word "and", "or". The invention adopts an inaccurate reasoning method based on credibility in a MYCIN system, and has the main idea that the truth value of a fuzzy proposition combination formula takes the minimum value of each sub-formula truth value, the truth value of a fuzzy proposition analysis formula takes the maximum value of each sub-formula truth value, and the total number of the fuzzy proposition analysis formula is 5 types.

(1) Type "one in one fruit": IF p₁ Then p₂，t^·＝{p₂The transition enables to satisfy α₁W μ > λ, is triggered and the confidence of the fault event after triggering, α₂＝α₁w μ, as shown in fig. 3(a) -3 (b), where w is 1. Failure incidence rate f (p)₁)＝α₁μ。

(2) Type "multi-cause one fruit": IF p₁ AND p₂ AND…AND p_n THEN p_k，t^·＝{p_kThe transition enables to satisfy min (alpha)₁w₁μ，α₂w₂μ，…，α_nw_nμ) > λ, is triggered and a post-trigger fault event confidence α_k＝min(α₁w₁μ，α₂w₂μ，…α_nw_nMu), which is also called "condition and" mode, as shown in FIGS. 4(a) -4 (b), wherein

Failure incidence rate f (p)_1，2，...n)＝min(α₁μ，α₂μ，…α_nμ)。

(3) Type "one-cause-more-fruits": IF p_j THEN p₁ AND p₂ AND…AND p_n，t^·＝{p₁，p₂，…，p_np₁，p₂，...，p_nThe transition enables to satisfy α_jW μ > λ is triggered and the confidence of the fault event after triggering α₁＝α₂＝…＝α_n＝α_jW μ, this type is also called "conclusion and" pattern, as shown in fig. 5(a) -5 (b), where w is 1. Failure incidence rate f (p)_j)＝α_jμ。

(4) The "contention mode" type: IF p₁ AND p₂ AND…AND p_n THEN p_k，

If the transition enable satisfies max (alpha)₁w₁μ₁，α₂w₂μ₂，…，α_nw_nμ_n) Is triggered if lambda is greater than lambda, namely the confidence of fault event alpha after triggering_k＝max(α₁w₁μ₁，α₂w₂μ₂，…，α_nw_nμ_n) This type is also called "condition or" mode ", as shown in FIGS. 6(a) -6 (b), where w₁＝w₂＝…＝w_n1. Failure incidence rate f (p)₁)＝α₁μ₁，f(p₂)＝α₂μ₂，…，f(p_n)＝α_nμ_n。

(5) Type "multifactorial and multifruit": IF p_j1 AND p_j2 AND…AND p_jm THEN p_k1 AND p_k2 AND…AND p_knIf the transition enable satisfies min (. alpha.)_j1w₁μ，α_j2w₂μ，…，α_jmw_mMu) > lambda is triggered, i.e. the confidence of the fault event after triggering, alpha_k1＝α_k2＝…＝α_kn＝min(α_j1w₁μ，α_j2w₂μ，…α_jmw_mμ). Failure incidence rate f (p)_{j1，j2，...jm})＝min(α_j1μ，α_j2μ，…，α_jmμ) as shown in fig. 7(a) to 7 (b).

According to the characteristics of fault propagation, fault information flows in the fault Petri network, and after the transition occurs, the number of the Tokens in the library does not change, but a new Token is generated in the output library of the transition. In this case, according to the conventional Petri net theory, the transition will occur endlessly and repeatedly, which violates the propagation characteristics of the fault. For this purpose, in a failure Perti network, a set of transition states T is introduced_tWhen a transition occurs, the corresponding element value is set to 1, which indicates that the fault has propagated along the path, so the transition should not occur any more.

The Petri net has the hierarchical characteristic, and the hierarchical problem of the fault can be well solved by utilizing the hierarchical characteristic. For an armored vehicle power plant, the system can be divided into a plurality of subsystems according to a hierarchical concept, and each subsystem is modeled by utilizing a fault Petri network. And then, the target library of the low-level fault Petri net is used as an input library of the high-level fault Petri net, so that the problem of the hierarchy of the complex system is solved. For delayed nature of the failure, the transition or library may be associated with a function of time.

The step 3) is specifically as follows:

firstly, a fault propagation process is described through reasoning on event confidence coefficient and reasoning on transition ignition and fault events, forward reasoning is completed, and the fault state is evaluated.

(1) Inference of event confidence

As shown in fig. 8, through a modeling rule, a MYCIN confidence matrix reasoning method is adopted to obtain all state quantities of system elements through stepwise reasoning, and for convenience of reasoning, a weight element corresponding to a target library is particularly specified to be 1.

3 special operators are defined:

operator for taking large value

A, B and C are all m × n matrix, then C_ij＝max(α_ij，b_ij) I is 1, 2, …, m, j is 1, 2, …, n. Namely, it is

Multiplication operator

A, B and C are the matrix of m × q, q × n and m × n respectively, then

i＝1，2，…，m，j＝1，2，…，n。

Direct multiplier operator: c is A, B, C is m-dimensional vector, then C_i＝a_i·b_i，i＝1，2，…，m。

Thus, the confidence inference formula for an event can be derived as:

the reasoning process is as follows:

the first step is as follows: initializing parameters;

the second step is that: k is 0;

the third step: according to formula (4) from alpha^kCalculating to obtain alpha^k+1；

The fourth step: if α is^k+1≠α^kRepeating the third step; if α is^k+1＝α^kAnd the reasoning is finished.

(2) Transition firing and fault event reasoning

The inference formula of the transition support degree is as follows according to the modeling rule

By definition, when

When the temperature of the water is higher than the set temperature,

time of flight_i，

The transition-enabled conditions include:

condition 1:

middle element

Confidence threshold vector element λ greater than corresponding lower-level fault causing higher-level fault_j；

Condition 2:

greater than f_i。

By definition, T^k+1 _t＝F(D^k，λ，M^k)

Wherein, { m_iIs as^·T_jMinimal cut set { p_iIdentification of failure events.

Define 1 special operator:

comparison operator [ ]: c ═ A ^ B, A, B, C are m-dimensional vectors, then

For the race type, if a transition is enabled, to prevent repetitive firing, other transitions that are in race with it are considered enabled at the next inference whether or not conditions are met.

Triggered transition vector T_tHas the following reasoning formula

The inference formula of the fault event vector M is

Event confidence after ignition for forward propagation is

Wherein alpha is_fire ⁰＝α⁰。

Therefore, the inference process of transition firing is:

the first step is as follows: initializing parameters;

the second step is that: calculating the transition support degree according to the formula (7);

the third step: calculating the triggered transition vector T according to equation (9)_t；

The fourth step: calculating a fault event vector M according to equation (10);

the fifth step: if M is^k+1≠M^kRepeating the second step; if M is^k+1＝M^kAnd the reasoning is finished.

Then, the fault mode/phenomenon of a certain system level is observed/monitored to explore the root cause of the fault, namely, a minimal cut set causing the fault phenomenon is found, reverse reasoning is completed, and the reasoning path is determined by the fault susceptibility.

Definition 4: let t_aIs a transition, p_i，p_j，p_kThree libraries if p_i∈^·t_a，p_j∈t_a ^·Then p is_iIs referred to as p_jTo go back to the associated repository, p_jIs referred to as p_iTo the immediate look-ahead associative base. If p is_iIs p_jTo go back to the associated repository, p_jIs p_kThe immediate backtracking of the associated library is called p_iIs p_kThe backtracking associated library. If p is_i∈^·t_aAnd p is_k∈^·t_aThen is called p_iAnd p_kTo transition t_aAdjacent to the library. Comprising p_iImmediate backtracking association ofThe set of libraries is called p_iIs collected as IHS (p)_i) (ii) a Comprising p_iThe set of backtracking associative libraries is called p_iIs collected as HS (p)_i)。

And (3) reverse reasoning process:

the first step is as follows: initializing parameters;

the second step is that: finding a terminating node p_jIt is used as the initial target repository (p)_j，IHS(p_j) Whereinsaid: p is a radical of_jIs the place of the target library,^·p_j＝t_j，IHS(p_j) Is the target depot p_jTo compute an initial target library IHS (p)_j) Failure incidence of all elements;

the third step: searching a library p corresponding to the maximum fault susceptibility in all unsearched paths_kIt is used as the initial target repository (p)_k，IHS(p_k)). If IHS (p)_k) If the node is an empty set, the node is a termination node, the node returns to the immediate look-ahead association library, the third step is repeated, and otherwise the next step is carried out;

the fourth step: if IHS (p)_k) And if not, proceeding downwards according to the method of the third step until all the libraries are traversed.

And finally, combining reverse reasoning and forward reasoning to realize an intelligent diagnosis process.

The core idea of the forward and reverse reasoning intelligent diagnosis process is that the reverse reasoning is utilized to obtain an optimal search path, then the fault confidence coefficient of a bottom event is updated through sensing monitoring data, transition ignition judgment is carried out through the forward reasoning, and finally a fault propagation path is obtained. If the propagation path of the target event is obtained, reasoning is finished, and the bottom layer fault is considered to be the cause of the target fault; if the propagation path is not connected, the reverse reasoning is continued until the propagation path is found, and the flow is shown as 9.

Forward and reverse reasoning intelligent diagnosis process:

the first step is as follows: initializing parameters;

the second step is that: starting from a target library, obtaining a first path to a termination node through reverse reasoning;

the third step: and judging whether the path minimum cut set can be transmitted to the target library or not according to the transition ignition condition through forward transmission. If the path can reach, the fault source is found, namely the path minimal cut set is obtained; otherwise, the step goes to the second step until the fault source is found.

According to the invention, a system-level fault model is established according to the hierarchy and the correlation of the fault. And then, aiming at the characteristics of multi-level, multi-relevance, uncertainty and the like of the armored vehicle diesel engine fault, defining a fuzzy fault Petri network by taking a fuzzy theory and a Petri network theory as cores, determining a modeling rule and carrying out Petri network modeling on the armored vehicle power device. Finally, a forward reasoning algorithm is proposed to obtain the confidence coefficient of the basic event, the transition ignition sequence and the fault propagation path; a reverse reasoning algorithm is proposed to obtain a search path of the fault; and a forward and backward reasoning intelligent diagnosis algorithm is provided to realize the diagnosis of system-level faults. The method is simple and easy to implement, breaks through the non-deterministic fault propagation modeling technology of complex equipment, and solves the difficult problems of characterization and description of a fault model.

In a preferred embodiment of the method of the invention, in a first step (S100), the power system is an armored vehicle power plant comprising the following components: the system comprises a crank connecting rod mechanism, a linkage mechanism, a gas distribution mechanism, a turbocharger, a cooling system, a lubricating system, an air supply system, a fuel supply system, a heating system, an exhaust system and a starting system, wherein the correlation relation is obtained according to the hierarchy and the relativity, the hierarchy is the longitudinal direction of faults, high-level faults can be caused by low-level faults, and the low-level faults can certainly cause the high-level faults; the correlation is the "transversal" of the fault, and is a system composed of a plurality of interconnected subsystems, and the fault of the subsystem is caused by the fault propagation of the subsystem related to the fault or the next-stage subsystem.

In a preferred embodiment of the method according to the invention, in the second step (S200), when p is₁Then p₂Occurrence of t^·＝{p₂The transition enables to satisfy α₁W μ > λ, is triggered and after triggering is thereforeConfidence of barrier event alpha₂＝α₁wμ。

In a preferred embodiment of the method according to the invention, in the second step (S200), when p is₁And p₂… and p_nGeneration, p_kOccurrence of t^·＝{p_kThe transition enables to satisfy min (alpha)₁w₁μ，α₂w₂μ，…，α_nw_nμ) > λ, is triggered and a post-trigger fault event confidence α_k＝min(α₁w₁μ，α₂w₂μ，…α_nw_nμ)。

In a preferred embodiment of the method according to the invention, in the second step (S200), when p is_jP after occurrence₁And p₂… and p_nOccurrence of t^·＝{p₁，p₂，…，p_nThe transition enables to satisfy α_jW μ > λ is triggered and the confidence of the fault event after triggering α₁＝α₂＝…＝α_n＝α_j*w*μ。

In a preferred embodiment of the method according to the invention, in the second step (S200), when p is₁And p₂… and p_nP after occurrence_kThe occurrence of this is that,

if the transition enable satisfies max (alpha)₁w₁μ₁，α₂w₂μ₂，…，α_nw_nμ_n) Is triggered if lambda is greater than lambda, namely the confidence of fault event alpha after triggering_k＝max(α₁w₁μ₁，α₂w₂μ₂，…，α_nw_nμ_n)。

In a preferred embodiment of the method according to the invention, in the second step (S200), when p is_j1And p_j2… and p_jmP after occurrence_k1And p_k2… and p_knOccurs if the transition enables min (alpha)_j1w₁μ，α_j2w₂μ，…，α_jmw_mMu) > lambda is triggered, i.e. the confidence of the fault event after triggering, alpha_k1＝α_k2＝…＝α_kn＝min(α_j1w₁μ，α_j2w₂μ，…α_jmw_mμ)。

In the preferred embodiment of the method of the present invention, in the third step (S300), a MYCIN confidence matrix reasoning method is adopted to gradually infer all state quantities of system elements, and 3 special operators are defined: operator for taking large value

A, B and C are all m × n matrix, then C_ij＝max(a_ij，b_ij) I 1, 2, …, m, j 1, 2, …, n, multiplier

A, B and C are the matrix of m × q, q × n and m × n respectively, then

Direct multiplier operator: c is A, B, C is m-dimensional vector, then C_i＝α_i·b_i，i＝1，2，…，m

The confidence inference formula for an event is:

the reasoning process is as follows:

the first step is as follows: initializing parameters;

the second step is that: k is 0;

the third step: from alpha^kCalculating to obtain alpha^k+1；

The fourth step: if α is^k+1≠α^kRepeating the third step; if α is^k+1＝α^kAnd the reasoning is finished.

In a preferred embodiment of the method according to the present invention, in the third step (S300), the inference process of transition firing is:

the first step is as follows: initializing parameters;

the second step is that: according to

Calculating the transition support degree;

the third step: according to

Calculating the triggered transition vector T_t；

The fourth step: according to

Calculating a fault event vector M;

the fifth step: if M is^k+1≠M^kRepeating the second step; if M is^k+1＝M^kAnd the reasoning is finished.

In a preferred embodiment of the process according to the invention, t_aIs a transition, p_i，p_j，p_kThree libraries if p_i∈^·t_a，p_j∈t_a ^·Then p is_iIs referred to as p_jTo go back to the associated repository, p_jIs referred to as p_iIf p is an immediate look-ahead associative base of_iIs p_jTo go back to the associated repository, p_jIs p_kThe immediate backtracking of the associated library is called p_iIs p_kIf p is a backtracking relationship of_i∈^·t_aAnd p is_k∈^·t_aThen is called p_iAnd p_kTo transition t_aThe adjacent library of (1) contains p_iThe set of immediate backtracking associative banks is called p_iIs collected as IHS (p)_i) (ii) a Comprising p_iThe set of backtracking associative libraries is called p_iIs collected as HS (p)_i) And the reverse reasoning process:

the first step is as follows: initializing parameters;

the second step is that: finding a terminating node p_jIt is used as the initial target repository (p)_j，IHS(p_j) Whereinsaid: p is a radical of_jIs the place of the target library,^·p_j＝t_j，IHS(p_j) Is the target depot p_jTo compute an initial target library IHS (p)_j) Failure incidence of all elements;

the third step: searching a library p corresponding to the maximum fault susceptibility in all unsearched paths_kIt is used as the initial target repository (p)_k，IHS(p_k) If IHS (p)_k) If the node is an empty set, the node is a termination node, the node returns to the immediate look-ahead association library, the third step is repeated, and otherwise the next step is carried out;

the fourth step: if IHS (p)_k) And if not, proceeding downwards according to the method of the third step until all the libraries are traversed.

For further understanding of the present invention, the use of the present invention is illustrated in the diagnostic example of excessive oil temperature.

Firstly, according to the step 1), collecting and counting failure modes of subsystems or components causing the failure of overhigh engine oil temperature, and obtaining a failure level of a component-subsystem-system level by combining the subsystem structure of the armored vehicle power device obtained by analysis in the step one.

Then according to step 2), establishing a fuzzy fault Petri net model according to the definition of the fuzzy fault Petri net and the modeling rule, such as

As shown in the figure. The failure events in the model are shown in table 3.

TABLE 3 Engine oil over-temperature fault event data sheet

Serial number	Storehouse	Event name
			1.	p008	Excessive engine oil temperature
2.	p100	Crank-link mechanism failure
			3.	p500	Cooling system failure
4.	p600	Lubrication system failure
			5.	p610	Oil pump
6.	p620	Fault of engine oil heat exchanger
			7.	p630	Fault of engine oil filter
8.	p611	Pump cavity or gear damage of engine oil pump
			9.	p621	Engine oilLeakage of heat exchanger
10.	p622	Reduction in heat dissipation efficiency of engine oil heat exchanger
			11.	p631	Clogging or breakdown of oil filter element
12.	p510	Water pump failure
			13.	p511	Damage of water pump impeller
14.	p513	The impeller of the water pump can not rotate
			15.	p130	Piston assembly failure
16.	p131	Piston top cracking and cylinder drawing
			17.	p133	Piston ablation

And finally, respectively carrying out forward reasoning, reverse reasoning and forward and reverse reasoning.

The method comprises the following specific steps:

1) forward reasoning

Setting the confidence of the fault input by the initial library:

α_IN611＝0.2；α_IN621＝0.21；α_IN622＝0.23；α_IN631＝0.25；α_IN131＝0.98；α_IN133＝0.65；α_IN511＝0.14；α_IN513＝0.24；

default fault confidence:

α_default100＝0.05；α_default500＝0.07；α_default600＝0.06；α_default130＝0.05；α_default131＝0.04；α_default133＝0.08；α_default510＝0.07；α_default511＝0.08；α_default513＝0.07；α_default610＝0.06；α_default620＝0.05；α_default630＝0.04；α_default611＝0.07；α_default621＝0.06；α_default622＝0.05；α_default631＝0.04；

migration rule confidence

μ₁₀₀＝0.98；μ₅₀₀＝0.96；μ₆₀₀＝0.95；μ₆₁₀＝1；μ₆₂₀＝1；μ₆₃₀＝1；μ₆₁₁＝0.98；μ₆₂₁＝0.98；μ₆₂₂＝0.97；μ₆₃₁＝0.98；μ₁₃₀＝1；μ₁₃₁＝0.99；μ₁₃₃＝0.98；μ₅₁₁＝0.98；μ₅₁₃＝0.98；μ₅₁₀＝1；

Confidence threshold for lower layer faults causing higher layer faults

λ₁₀₀＝0.6；λ₅₀₀＝0.6；λ₆₀₀＝0.5；λ₆₁₀＝0.5；λ₆₂₀＝0；λ₆₃₀＝0；λ₆₁₁＝0.4；λ₆₂₁＝0.5；λ₆₂₂＝0.4；λ₆₃₁＝0.4；λ₁₃₀＝0；λ₁₃₁＝0.5；λ₁₃₃＝0.4；λ₅₁₁＝0.5；λ₅₁₃＝0.5；λ₅₁₀＝0；

Fuzzy probability threshold of library fault event

f₁₀₀＝0.4；f₅₀₀＝0.5；f₆₀₀＝0.5；f₁₃₀＝0.5；f₁₃₁＝0.5；f₁₃₃＝0.5；f₅₁₀＝0.5；f₅₁₁＝0.5；f₅₁₃＝0.5；f₆₁₀＝0.6；f₆₂₀＝0.5；f₆₃₀＝0.4；f₆₁₁＝0.5；f₆₂₁＝0.6；f₆₂₂＝0.5；f₆₃₁＝0.5；f₀₀₈＝0.7

An inference of an event confidence vector can be obtained using equation (6):

first propagation:

α₆₁₁＝0.2；α₆₂₁＝0.21；α₆₂₂＝0.23；α₆₃₁＝0.25；α₁₃₁＝0.98；α₁₃₃＝0.65；α₅₁₁＝0.14；α₅₁₃＝0.24；α₆₁₀＝0.196；α₆₂₀＝0.2231；α₆₃₀＝0.245；α₁₃₀＝0.9702；α₅₁₀＝0.2352；α₆₀₀＝0；α₁₀₀＝0；α₅₀₀＝0；α₀₀₈＝0

and (3) second propagation:

α₆₁₁＝0.2；α₆₂₁＝0.21；α₆₂₂＝0.23；α₆₃₁＝0.25；α₁₃₁＝0.98；α₁₃₃＝0.65；α₅₁₁＝0.14；α₅₁₃＝0.24；α₆₁₀＝0.196；α₆₂₀＝0.2231；α₆₃₀＝0.245；α₁₃₀＝0.9702；α₅₁₀＝0.2352；α₆₀₀＝0.245；α₁₀₀＝0.9702；α₅₀₀＝0.2352；α₀₀₈＝0

and (3) third propagation:

α₆₁₁＝0.2；α₆₂₁＝0.21；α₆₂₂＝0.23；α₆₃₁＝0.25；α₁₃₁＝0.98；α₁₃₃＝0.65；α₅₁₁＝0.14；α₅₁₃＝0.24；α₆₁₀＝0.196；α₆₂₀＝0.2231；α₆₃₀＝0.245；α₁₃₀＝0.9702；α₅₁₀＝0.2352；α₆₀₀＝0.245；α₁₀₀＝0.9702；α₅₀₀＝0.2352；α₀₀₈＝0.9508

fourth propagation:

α₆₁₁＝0.2；α₆₂₁＝0.21；α₆₂₂＝0.23；α₆₃₁＝0.25；α₁₃₁＝0.98；α₁₃₃＝0.65；α₅₁₁＝0.14；α₅₁₃＝0.24；α₆₁₀＝0.196；α₆₂₀＝0.2231；α₆₃₀＝0.245；α₁₃₀＝0.9702；α₅₁₀＝0.2352；α₆₀₀＝0.245；α₁₀₀＝0.9702；α₅₀₀＝0.2352；α₀₀₈＝0.9508

and the third and fourth reasoning results are the same, and the reasoning is finished.

The inference of the triggered transition vector can be obtained by using equation (9):

first propagation:

T_{t611_610}＝0；T_{t621_620}＝0；T_t622-620＝0；T_t631-630＝0；T_t131-130＝1；T_t133-130＝1；T_t511-510＝0；T_t513-510＝0；T_t610-600＝0；T_t620-600＝0；T_t630-600＝0；T_t130-100＝0；T_t510-500＝0；T_t600-008＝0；T_t100-008＝0；T_t500-008＝0

and (3) second propagation:

T_{t611_610}＝0；T_{t621_620}＝0；T_t622-620＝0；T_t631-630＝0；T_t131-130＝1；T_t133-130＝1；T_t511-510＝0；T_t513-510＝0；T_t610-600＝0；T_t620-600＝0；T_t630-600＝0；T_t130-100＝1；T_t510-500＝0；T_t600-008＝0；T_t100-008＝0；T_t500-008＝0

and (3) third propagation:

T_{t611_610}＝0；T_{t621_620}＝0；T_t622-620＝0；T_t631-630＝0；T_t131-130＝1；T_t133-130＝1；T_t511-510＝0；T_t513-510＝0；T_t610-600＝0；T_t620-600＝0；T_t630-600＝0；T_t130-100＝1；T_t510-500＝0；T_t600-008＝0；T_t100-008＝1；T_t500-008＝0

fourth propagation:

T_{t611_610}＝0；T_{t621_620}＝0；T_t622-620＝0；T_t631-630＝0；T_t131-130＝1；T_t133-130＝1；T_t511-510＝0；T_t513-510＝0；T_t610-600＝0；T_t620-600＝0；T_t630-600＝0；T_t130-100＝1；T_t510-500＝0；T_t600-008＝1；T_t100-008＝1；T_t500-008＝1

fifth propagation:

T_{t611_610}＝0；T_{t621_620}＝0；T_t622-620＝0；T_t631-630＝0；T_t131-130＝1；T_t133-130＝1；T_t511-510＝0；T_t513-510＝0；T_t610-600＝0；T_t620-₆₀₀＝0；T_t630-600＝0；T_t130-100＝1；T_t510-500＝0；T_t600-008＝1；T_t100-008＝1；T_t500-008＝1

and the fourth and fifth reasoning results are the same, and the reasoning is finished.

The inference of the fault event vector can be obtained using equation (10):

first propagation:

M₆₁₁＝0；M₆₂₁＝0；M₆₂₂＝0；M₆₃₁＝0；M₁₃₁＝1；M₁₃₃＝0；M₅₁₁＝0；M₅₁₃＝0；M₆₁₀＝0；M₆₂₀＝0；M₆₃₀＝0；M₁₃₀＝1；M₅₁₀＝0；M₆₀₀＝0；M₁₀₀＝0；M₅₀₀＝0；M₀₀₈＝0

and (3) second propagation:

M₆₁₁＝0；M₆₂₁＝0；M₆₂₂＝0；M₆₃₁＝0；M₁₃₁＝1；M₁₃₃＝0；M₅₁₁＝0；M₅₁₃＝0；M₆₁₀＝0；M₆₂₀＝0；M₆₃₀＝0；M₁₃₀＝1；M₅₁₀＝0；M₆₀₀＝0；M₁₀₀＝1；M₅₀₀＝0；M₀₀₈＝0

and (3) third propagation:

M₆₁₁＝0；M₆₂₁＝0；M₆₂₂＝0；M₆₃₁＝0；M₁₃₁＝1；M₁₃₃＝0；M₅₁₁＝0；M₅₁₃＝0；M₆₁₀＝0；M₆₂₀＝0；M₆₃₀＝0；M₁₃₀＝1；M₅₁₀＝0；M₆₀₀＝0；M₁₀₀＝1；M₅₀₀＝0；M₀₀₈＝1

fourth propagation:

M₆₁₁＝0；M₆₂₁＝0；M₆₂₂＝0；M₆₃₁＝0；M₁₃₁＝1；M₁₃₃＝0；M₅₁₁＝0；M₅₁₃＝0；M₆₁₀＝0；M₆₂₀＝0；M₆₃₀＝0；M₁₃₀＝1；M₅₁₀＝0；M₆₀₀＝0；M₁₀₀＝1；M₅₀₀＝0；M₀₀₈＝1

and the third and fourth reasoning results are the same, and the reasoning is finished.

The inference of the firing event confidence vector can be obtained using equation (11):

first propagation:

α_fire611＝0.2；α_fire621＝0.21；α_fire622＝0.23；α_fire631＝0.25；α_fire131＝0.98；α_fire133＝0.65；α_fire511＝0.14；α_fire513＝0.24；α_fire610＝0；α_fire620＝0；α_fire630＝0；α_fire130＝0.9702；α_fire510＝0；α_fire600＝0；α_fire100＝0；α_fire500＝0；α_fire008＝0

and (3) second propagation:

α_fire611＝0.2；α_fire621＝0.21；α_fire622＝0.23；α_fire631＝0.25；α_fire131＝0.98；α_fire133＝0.65；α_fire511＝0.14；α_fire513＝0.24；α_fire610＝0；α_fire620＝0；α_fire630＝0；α_fire130＝0.9702；α_fire510＝0；α_fire600＝0；α_fire100＝0.9702；α_fire500＝0；α_fire008＝0

and (3) third propagation:

α_fire611＝0.2；α_fire621＝0.21；α_fire622＝0.23；α_fire631＝0.25；α_fire131＝0.98；α_fire133＝0.65；α_fire511＝0.14；α_fire513＝0.24；α_fire610＝0；α_fire620＝0；α_fire630＝0；α_fire130＝0.9702；α_fire510＝0；α_fire600＝0；α_fire100＝0.9702；α_fire500＝0；α_fire0080.9508 fourth transmission:

α_fire611＝0.2；α_fire621＝0.21；α_fire622＝0.23；α_fire631＝0.25；α_fire131＝0.98；α_fire133＝0.65；α_fire511＝0.14；α_fire513＝0.24；α_fire610＝0；α_fire620＝0；α_fire630＝0；α_fire130＝0.9702；α_fire510＝0；α_fire600＝0；α_fire100＝0.9702；α_fire500＝0；α_fire008＝0.9508

and the third and fourth reasoning results are the same, and the reasoning is finished.

2) Reverse reasoning

Setting the failure incidence rate of the library:

α_k100＝0.05；α_k500＝0.07；α_k600＝0.06；α_k130＝0.05；α_k131＝0.04；α_k133＝0.08；α_k510＝0.07；α_k511＝0.08；α_k513＝0.07；α_k610＝0.06；α_k620＝0.05；α_k630＝0.04；α_k611＝0.07；α_k621＝0.06；α_k622＝0.05；α_k631＝0.04；

from the initial target repository P₀₀₈From this, it immediately traces back the IHS (p) aggregated by the associative base₀₀₈) Is { P₁₀₀，P₅₀₀，P₆₀₀Due to failure incidence rate alpha_k500＞α_k600＞α_k100Searching the path to P in reverse₅₀₀. Then it is used as the initial target library and its immediate backtracking associated library set IHS (p)₅₀₀) Is { P₅₁₀Reverse search path to P₅₁₀. Then it is used as the initial target library and its immediate backtracking associated library set IHS (p)₅₀₀) Is { P₅₁₀}. Then it is used as the initial target library and its immediate backtracking associated library set IHS (p)₅₁₀) Is { P₅₁₁，P₅₁₃Due to failure incidence rate alpha_k511＞α_k513Searching the path to P in reverse₅₁₁. Due to IHS (p)₅₁₁) For an empty set, the node is said to be a termination node, completing the first path (P)₀₀₈→P₅₀₀→P₅₁₀→P₅₁₁) To search for (1). Then, return its immediate look-ahead associative base P₅₁₀Searching again to obtain a second path (P)₀₀₈→P₅₀₀→P₅₁₀→P₅₁₃) To search for (1). And the rest is done by analogy to finish all searching paths.

3) Forward and reverse reasoning

According to the initialization conditions, a first search path (P) is first obtained by reverse reasoning₀₀₈→P₅₀₀→P₅₁₀→P₅₁₁). Then follows the forward reasoning process, following path (P)₅₁₁→P₅₁₀→P₅₀₀→P₀₀₈) Obtaining a transition firing sequence M₅₁₀＝0；M₅₀₀＝0；M₀₀₈If 1, indicating the failure of ignition, the search is reversed until the ignition path is found (P)₁₃₁→P₁₃₀→P₁₀₀→P₀₀₈)，M₁₃₁＝1；M₁₃₀＝1；M₁₀₀＝1；M₀₀₈The ignition condition is satisfied at 1. Therefore, the failure sources of piston top cracking and cylinder pulling are found.

Although the embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the above-described embodiments and application fields, and the above-described embodiments are illustrative, instructive, and not restrictive. Those skilled in the art, having the benefit of this disclosure, may effect numerous modifications thereto without departing from the scope of the invention as defined by the appended claims.

Claims (8)

1. A power system fault processing method based on a fuzzy fault Petri net comprises the following steps:

in the first step (S100), the association relationship between the components and the faults is established based on the fault type of the power system,

in a second step (S200), a fuzzy fault Petri net and a modeling rule are established, the fuzzy fault Petri net continuously changes the state of the system to describe the dynamic characteristics of the power system according to a transition enabling rule and an occurrence rule, wherein,

the fuzzy fault Petri net is defined as 12 tuples (P, T, I, O, K, T)_t，M，w，f，α，λ，U_u) Wherein: p ═ P₁，p₂，…，p_mIs a collection of non-empty finite libraries representing a collection of system failure events, T ═ T₁，t₂，…t_nThe method is characterized in that the method is a non-empty finite transition set, represents state change or behavior action of a simulated system, and reflects propagation evolution of internal faults of the system, and I: p × T is an input matrix, representing P_i→t_jThe existence of the directional arc between the two,namely transition T_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; o: t × P is an output matrix representing the output from T_j→p_iDirected arcs existing between, i.e. transitions t_jIs output arc of, and p_iTo transition t_jI is 1, 2, …, m; j is 1, 2, …, n; k: p → [0, 1]Is the capacity function of the fault Petri net; t is_t: triggered transition vector with initial value of T⁰＝(0，0，…，0)^TThe element is set to 1 after the same fault occurs, which means that the element cannot occur again before being repaired; m ═ M₁，m₂，…，m_m)^TDistribution vectors, m, identified for the library_iIdentify its corresponding depot p_iState of (1), m₁Is the initial vector identification of the library, and represents the initial state of the simulated system; w ═ w (w)₁，w₂，…，w_n)^TIs a library site fault event weight vector, reflects the input library site p_kDegree of influence on transition t, wherein

p_kIs e.g. I and

f＝(f₁，f₂，…，f_m)^Tis a fuzzy probability threshold vector of the fault event of the library, and the confidence coefficient alpha of the fault event_i＞f_iIf the fault event is true, otherwise, the fault event is false; α ═ α (α)₁，α₂，…，α_m)^TFor fault event confidence vectors, including α_default，α_INAnd alpha_OUT，α_defaultObtaining a default fault event confidence coefficient vector according to fault statistics, wherein the default fault event confidence coefficient vector comprises all the libraries; alpha is alpha_INObtaining a component fault diagnosis according to sensor monitoring data after inputting a fault event confidence coefficient vector, wherein the element is an initial library; alpha is alpha_OUTObtaining a confidence coefficient vector of the output fault event according to a forward reasoning algorithm, wherein elements are all the libraries except the initial library; λ ═ λ₁，λ₂，…，λ_n)^TA confidence threshold representing that a lower-level fault causes a higher-level fault; u shape_μ＝diag(μ₁，μ₂，…，μ_n) Denotes a transition confidence matrix, μ_jRepresents a transition t_jReliability of, mu_j∈[0，1]J 1, 2, …, N, defining a vector N_μ＝(μ₁，μ₂，…，μ_n)^TRepresents a transition confidence vector when p₁Generation then p₂Occurrence of t^·＝{p₂Denotes if library p₁Reach a repository p only through a unique transition t₂And the last set of transitions t is only { p }₂The transition enables to satisfy α₁W μ > λ, i.e. library p₁The actual confidence of the rule is greater than the threshold required for transition triggering, and the confidence of the fault event after triggering is alpha₂＝α₁wμ；

In the third step (S300), a fault event confidence coefficient, transition ignition and an event fault reasoning process are obtained based on a forward reasoning algorithm so as to obtain a basic event confidence coefficient, a transition ignition sequence and a fault propagation path; obtaining a search path of the fault based on a reverse reasoning algorithm to obtain a rapid diagnosis strategy of the fault; and diagnosing the system-level fault based on a forward and backward reasoning diagnosis algorithm.

2. The method of claim 1, wherein in a first step (S100), the power system is an armored vehicle power plant comprising the following components: the system comprises a crank connecting rod mechanism, a linkage mechanism, a gas distribution mechanism, a turbocharger, a cooling system, a lubricating system, an air supply system, a fuel supply system, a heating system, an exhaust system and a starting system, wherein the correlation relation is obtained according to the hierarchy and the relativity, the hierarchy is the longitudinal direction of faults, high-level faults can be caused by low-level faults, and the low-level faults can certainly cause the high-level faults; the correlation is the "transversal" of the fault, and is a system composed of a plurality of interconnected subsystems, and the fault of the subsystem is caused by the fault propagation of the subsystem related to the fault or the next-stage subsystem.

3. The method according to claim 1, wherein, in a second step (S200), when p is₁And p₂… and p_nGeneration, p_kOccurrence of t^·＝{p_k}, library site p₁The actual credibility of the rule is larger than the threshold value required by the transition trigger, and the transition enable meets min (alpha)₁w₁μ，α₂w₂μ，…，α_nw_nMu) > lambda, i.e. library p₁，p₂，…，p_nThe minimum of the actual confidence levels for the rules is greater than the threshold required for transition triggering, and the confidence level alpha of the fault event after triggering is triggered_k＝min(α₁w₁μ，α₂w₂μ，…α_nw_nμ)。

4. The method according to claim 1, wherein, in a second step (S200), when p is_jP after occurrence₁And p₂… and p_nOccurrence of t^·＝{p₁，p₂，…，p_nDenotes if library p₁，p₂，…，p_nAll can reach the library site p simultaneously through the transition t₁，p₂，…，p_nAnd the last set of transitions t is only { p }₁，p₂，…，p_nThe transition enables to satisfy α_jW μ > λ is triggered, i.e. library p₁，p₂，…，p_nThe minimum of the actual confidence levels for the rules is greater than the threshold required for transition triggering, and the confidence level alpha of the fault event after triggering is triggered₁＝α₂＝…＝α_n＝α_j*w*μ。

5. The method according to claim 1, wherein, in a second step (S200), when p is₁And p₂… and p_nP after occurrence_kThe occurrence of this is that,

i.e. depot p₁，p₂，…，p_nThe minimum of the actual confidence in the rule is greater than the threshold required for transition triggering, and the rule is triggered if the transition enable satisfies max (alpha)₁w₁μ₁，α₂w₂μ₂，…，α_nw_nμ_n) λ, i.e. library p₁，p₂，…，p_nThe maximum actual reliability of the rule is triggered as long as the maximum actual reliability is larger than the threshold required by the transition trigger, and the confidence alpha of the fault event after the trigger_k＝max(α₁w₁μ₁，α₂w₂μ₂，…，α_nw_nμ_n)。

6. The method according to claim 1, wherein in the third step (S300), the MYCIN confidence matrix reasoning method is used to deduce all state quantities of system components step by step, and 3 operators are defined: operator for taking large value

A, B and C are all m × n matrix, then C_ij＝max(a_ij，b_ij) I 1, 2, …, m, j 1, 2, …, n, multiplier

A, B and C are the matrix of m × q, q × n and m × n respectively, then

Direct multiplier operator: c is A, B, C is m-dimensional vector, then C_i＝a_i·b_i，i＝1，2，…，m

The confidence inference formula for an event is:

in the formula, alpha^k+1And alpha^kThe confidence degrees of the event at the k-th inference and the k + 1-th inference are respectively; o is an output matrix representing the output from t_j→p_iDirected arcs existing between, i.e. transitions t_jIs output arc of, and p_iTo transition t_jI is 1, 2, …, m; j is 1, 2, …, n; u shape_μRepresenting a transition confidence matrix, element mu_jRepresents a transition t_jReliability of, mu_j∈[0，1]，j＝1，2，…，n；I^TFor input matrix, representing from p_i→t_jDirected arcs existing between, i.e. transitions t_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; w is the weight vector of the failure event in the library, which reflects the p input into the library_kDegree of influence on transition t, wherein

p_kIs e.g. I and

is an m-dimensional vector, describes the confidence that the rule is true,

the reasoning process is as follows:

the first step is as follows: initializing parameters;

the second step is that: k is 0;

the third step: from alpha^kCalculating to obtain alpha^k+1；

The fourth step: if α is^k+1≠α^kRepeating the third step; if α is^k+1＝α^kAnd the reasoning is finished.

7. The method of claim 1, wherein in the third step (S300), the inference process of transition firing is:

the first step is as follows: initializing parameters;

the second step is that: according to

Calculating the transition support, wherein D^k+1，D^kRespectively identifying the potential enabling transition support degree of the event at the kth time and the kth +1 time; n is a radical of_μ＝(μ₁，μ₂，…，μ_n)^TRepresenting a transition confidence vector; alpha is alpha^kThe confidence of the k-th inference for the event; i is^TFor input matrix, representing from p_i→t_jDirected arcs existing between, i.e. transitions t_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; w is the weight vector of the failure event in the library, which reflects the p input into the library_kThe degree of influence on the transition t,

in the formula, T^k+1 _t，T^k _tRespectively identifying triggered transition vectors at the kth time and the kth +1 time; d^kIdentifying a support for the event at the kth potential enabling transition; λ ═ λ₁，λ₂，…，λ_n)^TA confidence threshold representing that a lower-level fault causes a higher-level fault; alpha is alpha^kThe confidence of the k-th inference for the event; i is^TFor input matrix, representing from p_i→t_jDirected arcs existing between, i.e. transitions t_jIs input arc of, and p_iTo transition t_jI ═ 1, 2, …, m; j is 1, 2, …, n; f. of^kIs a fuzzy probability threshold vector of the fault event of the kth inference bank;

the fourth step: according to

Calculating a fault event vector M, where M^k+1，M^kFault event vectors for the kth and the (k + 1) th times, respectively; t is^k _tIdentifying the triggered transition vector for the kth time; o is an output matrix representing the slave T_j→p_iDirected arcs existing between, i.e. transitions t_jIs output arc of, and p_iTo transition t_jI is 1, 2, …, m; j is 1, 2, …, n;

the fifth step: if M is^k+1≠M^kRepeating the second step; if M is^k+1＝M^kAnd the reasoning is finished.

8. The method of claim 1, wherein t is_aIs a transition, p_i，p_j，p_kThree libraries if p_i∈^·t_a，p_j∈t_a ^·Then p is_iIs referred to as p_jTo go back to the associated repository, p_jIs referred to as p_iIf p is an immediate look-ahead associative base of_iIs p_jTo go back to the associated repository, p_jIs p_kThe immediate backtracking of the associated library is called p_iIs p_kIf p is a backtracking relationship of_i∈^·t_aAnd p is_k∈^·t_aThen is called p_iAnd p_kTo transition t_aThe adjacent library of (1) contains p_iThe set of immediate backtracking associative banks is called p_iIs collected as IHS (p)_i) (ii) a Comprising p_iThe set of backtracking associative libraries is called p_iIs collected as HS (p)_i) And the reverse reasoning process:

the first step is as follows: initializing parameters;

the second step is that: finding a terminating node p_jIt is used as the initial target repository (p)_j，IHS(p_j) Whereinsaid: p is a radical of_jIs the place of the target library,^·p_j＝t_j，IHS(p_j) Is the target depot p_jTo compute an initial target library IHS (p)_j) Failure incidence of all elements;

the third step: searching a library p corresponding to the maximum fault susceptibility in all unsearched paths_kIt is used as the initial target repository (p)_k，IHS(p_k) If IHS (p)_k) If the node is an empty set, the node is a termination node, the node returns to the immediate look-ahead association library, the third step is repeated, and otherwise the next step is carried out;

the fourth step: if IHS (p)_k) And if not, proceeding downwards according to the method of the third step until all the libraries are traversed.

Images