CN112613116A - Petri net quantum Bayesian fault diagnosis method for liquid rocket engine starting stage - Google Patents

Abstract

The invention discloses a Petri network quantum Bayes fault diagnosis method in a starting stage of a liquid rocket engine. By establishing a quantum Bayesian network model related to the unobservable fault transition, and by utilizing the observable node triggering relation and the triggering probability of the unobservable nodes, quantum parameters are manually selected to calculate the triggering probability of the fault nodes, and the fault state of the system is judged. The method can be used for fault diagnosis of a discrete dynamic system with a considerable part, and has the advantage that the fault diagnosis of the non-considerable part can be accurately carried out.

Description

A Petri Net Quantum Bayesian Fault Diagnosis Method for Liquid Rocket Engine Startup

technical field

The invention relates to the field of fault diagnosis of discrete dynamic systems, in particular to a Petri net quantum Bayesian fault diagnosis method in the start-up stage of a liquid rocket engine.

Background technique

Thrust failures account for about a quarter of spacecraft, satellite attitude and orbit control system failures, and thruster failures can dramatically alter the behavior of the entire spacecraft and may even lead to the failure of the entire mission. Therefore, abnormal component behavior must be detected early. Fault detection and diagnosis of spacecraft often rely on hardware and sensor redundancy, but adding sensors to spacecraft is not always successful due to constraints such as weight and cost.

In order to overcome the problem of incomplete observation information and meet the real-time requirements, the diagnosis system generally adopts the diagnosis method based on the qualitative model.

At present, the existing fault diagnosis technology is Petri net, which is first used in fault identification and diagnosis, and is applied in power system, communication system and so on. By using an incremental-based algorithm, starting from the set of current possible states and their probabilities, a probabilistic model is used to propose a method for evaluating the probability of future failures. The partially observable stochastic Petri net, on the other hand, calculates the probability of time- and non-time-marked trajectories consistent with a given time-observed sequence, and diagnoses based on the probability of failure.

The invention creates and proposes a Petri net quantum Bayesian fault diagnosis method in the start-up phase of a liquid rocket engine. Due to the introduction of quantum interference, the quantum Bayesian network can greatly affect probability inference, especially when the uncertainty level of the network is very high. high (no evidence observed). By establishing a quantum Bayesian network model about the transition of unobservable faults, using the triggering relationship of observable nodes and the triggering probability of unobservable nodes, manually selecting quantum parameters to calculate the triggering probability of faulty nodes, and judging the fault state of the system.

SUMMARY OF THE INVENTION

In view of this, the purpose of the present invention is to provide a Petri net quantum Bayesian fault diagnosis method in the start-up phase of a liquid rocket engine, and the present invention can greatly improve the efficiency in the fault diagnosis process and the accuracy of the fault diagnosis in the start-up phase of the liquid rocket engine. sex.

In order to achieve the above-mentioned purpose, the present invention provides a Petri net quantum Bayesian fault diagnosis method in the start-up stage of a liquid rocket engine, which is characterized in that, comprising the following steps:

Step S1, establishing a Petri net model of the liquid rocket engine startup stage according to the working state of the liquid rocket engine startup stage;

Step S2, constructing the SCG graph of the Petri net model, traversing the path in the SCG graph that satisfies the observable transition trigger sequence, and judging whether the fault transition is included, and estimating the fault state of the liquid rocket engine;

Step S3, establishing a QBPN model, and estimating the failure probability of the liquid rocket engine through the QBPN model.

Further, the step 3 specifically includes:

Step S301, calculating the failure transition trigger probability according to the observable transition trigger state in the forward path;

Step S302, using the QB function to correct the fault transition trigger probability according to the observable transition trigger state of the backward path;

Step S303, taking the maximum trigger probability among all failure transitions as the failure probability of the liquid rocket engine.

The beneficial effects of the present invention are:

The present invention can perform fault diagnosis for a part of the discrete dynamic system that is observable, and has the advantage of being able to perform accurate fault diagnosis for the insignificant part. Because of the introduction of quantum interference, by establishing a quantum Bayesian network model about the transition of unobservable faults, using the triggering relationship of observable nodes and the triggering probability of unobservable nodes, and manually selecting quantum parameters to calculate the triggering probability of faulty nodes, the system fault can be judged state.

Description of drawings

Fig. 1 is the Petri net model of each key node in the working process of the liquid rocket engine start-up phase in Example 1.

FIG. 2 is the QBPN model corresponding to the fault transition t ₁₀ in the first embodiment.

Detailed ways

In order to make the purposes, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

It should be noted that, before formally explaining the embodiments of the present invention, there are two points that need to be explained in advance. First, the present invention is to predict possible failures in the startup phase of the liquid rocket engine. The working principle of the liquid rocket engine startup phase is high pressure helium. The gas enters the propellant storage tank and squeezes the propellant into the downstream combustion chamber to mix and ignite to generate thrust. The solenoid valve does not open or the fuel filling valve fails, indicating that the liquid rocket engine failed during the startup phase.

Second, two algorithms related to the present invention are introduced. The first algorithm is:

1. State estimation algorithm based on SCG

和可观变迁序列σ_o；Input: POPN model and possible failure transitions

and the observable transition sequence σ _o ;

Output: fault status of the system

Step 1. Initialize the root node C ₀ and mark node(C ₀ )=E, corresponding to the mark M ₀ , t _i ∈ A(M ₀ );

Step 2. When there is a node marked node(C _k )=E, select a new node C _k to mark node(C _k )=E, traverse any t _i ∈ A(M ₀ ), t _i triggers on the marker M _k M _q =M _k +C(·,t _i ) to establish a new node C _q marked as M _q , if there is the same node as C _q , then C _q node mark node(C _k )=Z otherwise C _q node mark node ( C _k )=E;

Step 3. Construct the SCG graph with the nodes marked E;

Step 4. Put all paths satisfying S(σ _o ) into the set;

则

如果

则

如果

则

Step 5. Traverse i=1,...,r, if

but

if

but

if

but

为每个集合分配一个故障状态；所有满足可观变迁序列σ_o的路径放在集合S(σ₀)；Pre∈(N)^m×n、Post∈(N)^m×n分别为变迁的前向弧矩阵和后向弧矩阵，C＝Post-Pre为POPN的关联矩阵，维数为m×n(N为非负整数集)，M₀为初始标识；ψ为量子贝叶斯表；t_i∈A(M₀)表示从M₀状态到其他状态的路径集合。in,

Assign a fault state to each set; all paths satisfying the observable transition sequence σ _o are placed in the set S(σ ₀ ); Pre∈(N) ^m×n , Post∈(N) ^m×n are the forward directions of the transition, respectively Arc matrix and backward arc matrix, C=Post-Pre is the correlation matrix of POPN, the dimension is m×n (N is the set of non-negative integers), M ₀ is the initial identification; ψ is the quantum Bayes table; t _i ∈A(M ₀ ) represents the set of paths from M ₀ state to other states.

In this algorithm, step 1, step 2, and step 3 are the process of constructing the SCG graph, step 4 is to traverse all the paths satisfying the observable transition sequence σ _o and put them in the set S(σ ₀ ), and step 5 is to traverse all the paths that satisfy the observable transition sequence σ o and put them in the set S(σ 0 ). Whether the path in σ ₀ ) contains fault transitions determines the fault state of the system.

2. Fault diagnosis algorithm based on QBPN model

Input: POPN model, quantum Bayesian probability table ψ, and possible failure transitions

Output: System Failure Probability

Step 1. Initialize

当

则i＜j≤k,t_j∈T_u，复制t_k与

之间所有变迁t_j、库所P和弧到QBPN，

当

则k＜j≤i,t_j∈T_u，复制

与t_k之间所有变迁t_j、库所P和弧到QBPN，

Step 2. Traverse

when

Then i＜j≤k,t _j ∈T _u , copy t _k and

All transitions between t _j , place P and arc to QBPN,

when

Then k＜j≤i,t _j ∈T _u , copy

All transitions t _j , places P and arcs to QBPN between and t _k ,

t_k没有发生权则删除t_k及其输入输出弧；Step 3. If

If t _k does not have the right to occur, delete t _k and its input and output arcs;

根据前向路径可观变迁状态计算

Step 4. Traverse t _{i ∈} T _QB , set P(t _i ) and quantum probability amplitude for t _i according to the quantum probability table ψ

Calculated according to the observable transition state of the forward path

令

计算系统故障概率

Step 5. Traverse

make

Calculate the probability of system failure

其中

为QBPN中可观变迁集，

为QBPN中不可观变迁集，且

ψ是t_i∈T_QB的量子概率表。Among them, the transition t _i is ignited at the mark M, denoted as M[t _i >, σ=t ₀ t ₁ t ₂ ... t _h ∈ T is the transition sequence set, and the corresponding mark is M ₀ [t ₀ >M ₁ [t ₁ >…[t _h >M _h+1 , abbreviated as M ₀ [σ>M _h+1 , t _h is the last triggered transition; T is the n-dimensional transition set, T _o is the observable transition set, T _u is the set of unobservable transitions; the quantum Bayesian Petri net is defined as

in

is the observable transition set in QBPN,

is the unobservable transition set in QBPN, and

ψ is the quantum probability table for t _{i ∈} T _QB .

In this algorithm: Step 1, Step 2, Step 3 and Step 4 are the process of constructing the QBPN model on the basis of the POPN model, and Step 2 is to construct the relational network containing the fault transition forward and backward respectively according to the sequence of the fault transition, Step 3 is to delete the transition without the right of occurrence, step 4 is to give the quantum probability amplitude triggered by the transition, and calculate the prior probability of the fault transition, and step 5 is to correct the trigger probability of the fault transition according to the observable transition trigger state of the backward path.

The QBPN model diagnoses the system fault state. First, the fault transition trigger probability is calculated according to the observable transition trigger state in the forward path, and then the QB function is used to correct the fault transition trigger probability according to the backward path observable transition trigger state.

Finally, the maximum trigger probability among all fault transitions is taken as the system fault probability.

Example 1

Figure 1 takes the key nodes in the starting stage of the liquid rocket engine as the warehouse, and the key action is the transition to establish a Petri net model to simulate the starting stage of the liquid rocket engine. The meanings of each warehouse and transition are shown in Table 1 and Table 2.

Table 1. Physical meaning of each warehouse

Treasury Library meaning Treasury Library meaning p₀ system ready p₈ Oxidant pipeline temperature up to standard p₁ 1 branch connected to helium p₉ The fuel filling valve works stably p₂ 2 branches are connected to helium p₁₀ Oxidant Fill Valve Enable p₃ Fuel tank builds up air pressure p₁₁ Preparing the combustion chamber p₄ Oxidant tank builds up air pressure p₁₂ Oxidant tank to be pressurized p₅ Fuel tank air pressure up to standard p₁₃ The fuel tank needs to be pressurized p₆ Oxidant storage tank air pressure up to standard p₁₄ Combustion chamber gas concentration is low p₇ Fuel line temperature up to standard p₁₅ Engine blowout enable

Table 2. Physical meaning and observability of each transition

Referring to Figure 1 , if the solenoid valve (transition t ₁₀ ) does not open or the fuel fill valve (transition t ₁₁ ) fails, the system fails.

The present embodiment proposes a Petri net quantum Bayesian fault diagnosis method in the start-up phase of a liquid rocket engine, including the following steps:

Step S1, establishing a Petri net model of the liquid rocket engine startup stage according to the working state of the liquid rocket engine startup stage;

Step S2, constructing the SCG graph of the Petri net model, traversing the path in the SCG graph that satisfies the observable transition trigger sequence, and judging whether the fault transition is included, and estimating the fault state of the liquid rocket engine;

Step S3, establishing a QBPN model, and diagnosing the fault state of the liquid rocket engine through the QBPN model;

Specifically, the step 3 specifically includes:

Step S301, calculating the failure transition trigger probability according to the observable transition trigger state in the forward path;

The preceding path refers to the path passed before the point of the obtained probability. In this embodiment, for example, the forward path at point t10 in FIG. ₂ is t3 and t5;

Step S302, using the QB function to correct the fault transition trigger probability according to the observable transition trigger state of the backward path;

The backward path refers to the path passed after the point of the obtained probability, in this embodiment, for example, the path from a certain point in FIG. 2 to the observable nodes 12 , 14 , and 15 .

The modification specifically refers to modifying the posterior probability because the known conditions change. In this embodiment, for example, the probability of t ₁₀ in FIG. 2 has been calculated, and then the triggering of t ₁₄ is observed, and then the posterior probability of t ₁₀ has been calculated. will change. Similarly, the posterior probability of t ₁₀ also changes in the case where the trigger at t ₁₅ is observed. So it needs to be corrected.

Step S303, taking the maximum trigger probability among all failure transitions as the failure probability of the liquid rocket engine.

More specifically, the specific process of step 2 in this implementation is the state estimation algorithm based on SCG proposed in the specific implementation manner of the present invention.

According to Figure 1, the sequence observed in the first test is μ ₀ μ ₄ , see Table 3 for details, calculate all TS path sets, that is, there are 3 transition sequence path sets, 3 normal transitions, and no fault occurs .

The observation sequence of the second test is μ ₀ μ ₄ μ ₃ μ ₁₄ μ ₁₅ μ ₁₂ . See Table 3 for details. There are 15 TS sets σ that satisfy the conditions, including 3 normal transitions σ and 12 fault transitions σ bar, both t ₁₀ and t ₁₁ have the potential to trigger a fault.

According to the state estimation algorithm based on SCG, the test diagnoses that the system failure states are all possible failures, and the possible failure transitions are t ₁₀ (electromagnetic valve failure) and t ₁₁ (fuel filling valve failure).

Then, the fault diagnosis algorithm based on the QBPN model is used to estimate the system failure probability. First, the algorithm is used to construct the QBPN model corresponding to the fault transition t ₁₀ and the fault transition t ₁₁ respectively, and then the respective trigger probabilities of the two fault transitions are calculated, and the maximum probability is taken as the system failure probability. .

In the partially observable Petri net model of the liquid rocket engine startup phase, given a fault transition, the transitions and locations in all its paths are searched forward and backward until an observable transition occurs. For the fault transition t ₁₀ , its path includes t ₃ t ₅ t ₁₀ t ₁₄ t ₁₂ , t ₃ t ₅ t ₁₀ t ₁₅ t ₁₂ , t ₃ t ₅ t ₁₀ t ₉ t ₈ t ₁₅ t ₁₂ , t ₃ t ₅ t ₁₀ t ₉ t ₈ t ₁₄ t ₁₅ t ₁₂ , where the observable transitions in the path t ₃ t ₅ t ₁₀ t ₁₄ t ₁₂ are t ₃ , t ₁₂ , t ₁₄ , and the unobservable transitions are t ₅ , t ₁₀ ; In the path t ₃ t ₅ t ₁₀ t ₁₅ t ₁₂ the observable transitions are t ₃ , t ₁₂ , t ₁₅ , and the invisible transitions are t ₅ , t ₁₀ ; in the path t ₃ t ₅ t ₁₀ t ₉ t ₈ t ₁₅ t ₁₂ The observable transitions are t ₃ , t ₁₂ , t ₁₅ , the unobservable transitions are t ₅ , t ₁₀ , t ₉ , t ₈ paths, and the observable transitions in t ₃ t ₅ t ₁₀ t ₉ t ₈ t ₁₄ t ₁₅ t ₁₂ are t ₃ , t ₁₂ , t ₁₄ , t ₁₅ , the invisible transitions are t ₅ , t ₁₀ , t ₉ , t ₈ . The QBPN model diagnosis sub-network corresponding to the leakage fault of the fuel filling valve at t ₁₀ is established as shown in Figure 2.

In order to use quantum Bayesian inference to calculate the failure probability of the system in the QBPN model, it is necessary to obtain the quantum conditional probability table of the transition and the triggering state of the observable transition. These information reflect the uncertainty of the fault process and the quantum interference effect.

In the POPN-based liquid rocket engine start-up stage fault diagnosis of the present invention, the simulation experiment result is a possible fault, and the fault diagnosis algorithm based on the QBPN model needs to be used to determine the system fault probability.

Table 3. Diagnosis results of fault diagnosis algorithm based on QBPN model

Table 3 is the diagnosis results of the fault diagnosis algorithm based on the QBPN model. In the test, the transition of the fault to be diagnosed is t ₁₀ and t ₁₁ , and the monitoring center receives the following observable transition trigger states: observable transition t ₃ , t ₁₂ , t ₁₄ , t ₁₅ triggers.

It should be pointed out that the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. Any person skilled in the art can easily think of changes within the technical scope disclosed by the present invention. and replacement, should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims (2)

1. A Petri net quantum Bayesian fault diagnosis method in a starting stage of a liquid rocket engine is characterized by comprising the following steps:

s1, establishing a Petri network model of the liquid rocket engine in the starting stage according to the working state of the liquid rocket engine in the starting stage;

s2, constructing an SCG (substation configuration graph) of the Petri network model, traversing paths which meet an observable transition trigger sequence in the SCG, judging whether fault transition is included, and preliminarily estimating the fault state of the liquid rocket engine;

and step S3, establishing a QBPN model, and estimating the fault probability of the liquid rocket engine through the QBPN model.

2. The Petri net quantum Bayesian fault diagnosis method for the liquid rocket engine starting stage according to claim 1, wherein the step 3 specifically comprises:

step S301, calculating a fault transition triggering probability according to an observable transition triggering state in a forward path;

step S302, correcting the fault transition triggering probability according to the observable transition triggering state of the backward path by utilizing a QB function;

and S303, taking the maximum trigger probability in all fault transitions as the fault probability of the liquid rocket engine.

Images