CN112270128B - A fault diagnosis method for hydraulic end of drilling pump based on dynamic fault tree - Google Patents

Abstract

The invention discloses a drilling pump hydraulic end fault diagnosis method based on a dynamic fault tree, and relates to the field of fault diagnosis. The invention aims to improve a fault diagnosis method of a discrete-time Bayesian network based on a dynamic fault tree. The core of the invention is to construct a mapping relation of 'rate parameter lambda-division number n', and the time division number n can be selected more scientifically by using a fault diagnosis method of a discrete time Bayesian network based on a dynamic fault tree; the method solves the problem of inaccurate diagnosis caused by the fact that researchers select the time division number n subjectively by proposing a mode of establishing the corresponding relation of the rate parameter lambda-the division number n, improves the reliability of fault diagnosis, and can provide reference for other similar complex machines when using the fault diagnosis method.

Description

A fault diagnosis method for hydraulic end of drilling pump based on dynamic fault tree

technical field

The invention relates to the field of fault diagnosis, in particular to an improved discrete-time Bayesian network fault diagnosis method based on a dynamic fault tree.

Background technique

The fault tree analysis method is based on the premise of fully and accurately grasping the system structure and principle, first determine the top event of the system, and then analyze it step by step according to the central event, and find out all the causes of the top event until the bottom event. It constitutes a top-down diagnostic process. In order to establish the logical relationship between the top event, the middle event and the bottom event in the fault tree, the fault tree is continuously improved. After determining the fault tree, use qualitative analysis and quantitative analysis to analyze the fault tree. However, the failure of modern engineering is usually accompanied by complex dynamic characteristics such as failure time and sequence. Simply using the conventional fault tree method, the obtained fault accuracy is low. In the 1990s, the dynamic fault tree analysis method was proposed, which has strong modeling and analysis capabilities for systems with dynamic fault logic. The difference between dynamic fault tree and conventional fault tree is that a series of dynamic logic gates are introduced to describe the timing rules and dynamic aging behavior of the system. At present, the algorithms for solving dynamic fault tree mainly include: dynamic fault tree analysis algorithm based on Markov chain, dynamic fault tree analysis algorithm based on Bayesian network, and approximate algorithm of dynamic fault tree based on trapezoidal formula. By comparison, the algorithm based on Bayesian network is more suitable for dynamic fault tree analysis. But this method has a shortcoming, the time division number n is determined by the researcher subjectively or only one division number n is determined. For complex machinery, the state degradation of components or systems is carried out gradually over time, and the degradation efficiency of different objects is not the same.

SUMMARY OF THE INVENTION

The purpose of the present invention is to improve the fault diagnosis method of discrete time Bayesian network based on dynamic fault tree. The core of the present invention is to construct the mapping relationship of "rate parameter λ--division number n". When using the fault diagnosis method of discrete-time Bayesian network based on dynamic fault tree, the time division number n can be selected more scientifically, so that the fault diagnosis can be improved. The results are more reliable.

The technical scheme of the present invention is a method for diagnosing the hydraulic end of a drilling pump based on a dynamic fault tree, the method comprising:

Step 1: Use the deductive reasoning method to establish the dynamic fault tree of the drilling pump hydraulic system:

Step 1.1: Take the failure of the hydraulic end of the drilling pump as the top event of the dynamic fault tree, take the failure of the sub-components in the main component in the hydraulic end as the intermediate event, and the failure of the components in the sub-component as the bottom event;

Step 1.2: According to the working principle and failure mechanism of the hydraulic end of the drilling pump, use the corresponding dynamic logic gate to connect the events at all levels with the top event, from the bottom event to the top event, use the dynamic logic gate to connect the events at all levels to obtain the drilling pump. Dynamic fault tree of the liquid end;

Step 2: Convert the dynamic fault tree into a Bayesian network model, and establish the corresponding relationship of "rate parameter λ--time division number n";

Step 2.1: Convert the dynamic fault tree into a Bayesian network model; the top event, middle event, and bottom event of the dynamic fault tree correspond to the root nodes, middle nodes, and leaf nodes of the Bayesian network model. Each logic in the dynamic fault tree The expression method of the gate in the Bayesian network model is:

AND gate:

Let X=[X ₁ ,X ₂ ,...,X _m ], where m is the number of input events of the AND gate, X _i ,i=1,2,...,m is the state variable of the input event, and the number of state combinations thereof is (N+1) ^m ; let Y be the state variable output by the AND gate, the state space of all variables is {1,2,...,N+1}, let k=max(X ₁ ,X ₂ ,..., X _m ), j is the judgment fixed value; the failure mechanism of the AND gate is that the output event occurs when all input events occur, then the output event should be at the maximum value of the state values of all events, therefore, in any state combination of input events, Y The conditional probability distribution of is:

OR gate:

The failure mechanism of the OR gate is that as long as one of the input events occurs, the output event occurs, so the state of the OR gate output event is the same as the minimum value of the state in the input event; let r=min(X ₁ ,X ₂ ,...,X _m ), then the conditional probability distribution of the OR gate output event is:

Priority AND gate:

Assuming that the input events are A and B, the output event is Y, and the state values are a, b, and y, respectively, the failure mechanism of the priority AND gate is that when both A and B fail, A fails before B, and event Y will fail. occurs, that is, when a<b, the output event Y is in the state b; otherwise, the output event Y is in the state N+1, which means that Y has not failed; then the conditional probability distribution of the output event Y of the priority AND gate is as follows:

When a < b,

When a≥b,

Sequence dependent gates:

Sequence-dependent gates force their input events to occur in a particular order and not fail in any other order. Sequential correlation gates are similar to priority AND gates in that they both represent the timing of basic events. The difference between them is that input events in sequential correlation gates cannot fail in any order; while priority AND gates can fail in any order, only in a specific order. will trigger the failure of its output event. Therefore, the sequential correlation gate can be changed to the form of multiple priority AND gates cascaded;

Function related gates:

Suppose the function-related gate has only one related input event A, the trigger event is Tr, and the output event is T; when Tr occurs, its related event A occurs, and output event T occurs; when A fails independently, output event T also occurs; An intermediate node A' is added between the input event and the output event to represent the total failure event of A failure caused by Tr triggering or the independent failure of A itself, p, q are the corresponding event states, then the function-related gate output The conditional probability distribution of the event is:

Step 2.2: Set the value range of n, use multiple sets of data collected in advance, and select different rate parameters λ to calculate the reliability of the system; compare the maximum difference ratio of the system reliability between each group of data, the rate parameter can be obtained λ should select the most suitable time division number n in different ranges, and establish the corresponding relationship of "rate parameter λ--division number n";

Step 3: Define the state of the nodes in the Bayesian network model;

Divide the time interval [0, t] composed of the entire task into n sub-intervals of equal length, and the length of each sub-interval is Δ=t/n, then the entire time axis [0, +∞) is divided into n+1 sub-intervals interval; when a component corresponding to a node A fails within the i-th time interval within the task time t, that is, A fails within the time interval [(i-1)Δ, iΔ], then node A is said to be in state i ; If A does not fail within task time t, that is, A fails within [t, ∞), then A is in state n+1;

Through the above definition, the state space of all nodes in the Bayesian network is obtained as the following time interval: [0,Δ],(Δ,2Δ],…,((n-1)Δ,nΔ],(nΔ, +∞), abbreviated as {1,2,…,n+1}, the failure time X of the system and components always corresponds to a certain interval i of n+1 intervals; the sum of the probabilities that the system is in the first n states is the failure rate of the system at task time t, and the probability that the system is in the n+1th state is the non-failure rate of the system at task time t;

Step 4: establish the probability distribution of all nodes according to the state of the nodes in the Bayesian network model, and complete the quantitative calculation of the dynamic fault tree fault diagnosis based on the Bayesian network;

Step 5: Reverse reasoning to obtain fault diagnosis results;

The conditional probability distribution of each node is expressed as P(X _i |pa(X _i )), which is used to express the quantitative relationship between child nodes and parent nodes; the prior probability distribution at a given root node and the non-root node’s Under the condition of conditional probability distribution, the joint probability distribution including all nodes is obtained, and then the target node is marginalized to obtain its marginal probability distribution; finally, the failure probability of each bottom event is calculated by the reverse reasoning of Bayesian network. , arrange according to the size of the failure probability, and output the diagnosis results.

Further, the top event in the step 1.1 is the hydraulic end failure, the middle event is: suction pipe failure, valve failure, piston failure, liquid cylinder failure, discharge pipe failure, air bag failure, safety valve failure, and the bottom event is: : The suction pipeline is not tightly sealed, the suction filter is blocked, the valve is severely worn, the valve cover is not tightened, the guide sleeve is stuck, the piston is severely worn, the piston nut is loose, the liquid cylinder enters the air, the cylinder head is not tightened, the cylinder head The gland is loose, the discharge filter cartridge is blocked, the discharge pipeline is blocked, the inflation joint is blocked, the inner bag of the air bag is ruptured, the shut-off valve is not tightly sealed, and the safety valve is improperly set;

The relationship between the events in step 1.2 is:

The suction pipeline is not tightly sealed, the suction filter screen is blocked and the door is connected to the suction pipe fault; the valve is seriously worn, the valve cover is not tightened, the guide sleeve is stuck through or the door is connected to the valve fault; the piston is seriously worn, and the piston nut Looseness is connected to the fault of the piston through the OR door; the cylinder enters the air, the cylinder head is not tightened, the cylinder head gland is loose and is connected to the fault of the cylinder through the sequence-related door; the discharge filter cartridge is blocked, the discharge line is blocked, and the or door is connected to the discharge pipe The fault is connected; the inflation joint is blocked, the inner bag of the air bag is ruptured, and the shut-off valve is not tightly sealed. It is connected to the air bag fault through the function-related door. The rupture of the inner bag is the trigger event input of the function-related door, and the air bag failure is the output of the function-related door; the improper setting of the safety valve is connected with the door and the safety valve fault.

Further, the specific method of the step 5 is:

In the discrete-time Bayesian network DTBN containing m nodes, the non-leaf nodes are represented by events U _i , where 1≤i≤m-1, and the occurrence interval of U _i is {[0,Δ],(Δ,2Δ] ,...,((n-1)Δ,nΔ],(T,+∞)}; if the top event U _T occurs within the task time T, then the top event must occur at [0,Δ], (Δ,2Δ],...,((n-1)Δ,nΔ],(nΔ,+∞) in one of the intervals; therefore, the probability of U _T occurring within the task time T can be directly calculated and expressed for:

In the formula, _ui represents the occurrence interval of U _i , and _ui belongs to {[0,Δ],(Δ,2Δ],...,((n-1)Δ,nΔ],(T,+∞)} ;

Then the failure probability of each component can be obtained when the system fails by using the method of reverse reasoning.

Advantages of the present invention: The present invention eliminates the problem of inaccurate diagnosis caused by the researcher's subjective selection of the time division number n by proposing to establish a corresponding relationship between "rate parameter λ--division number n", and improves the reliability of fault diagnosis. , and can provide a reference for other similar complex machinery using this fault diagnosis method.

Description of drawings

Fig. 1 is a flow chart of the present invention.

Figure 2 shows the dynamic fault tree model of the hydraulic end of the drilling pump.

Figure 3 is a system reliability comparison curve when n takes different values.

Figure 4 is a diagram of the conversion of an AND gate/OR gate/precedence AND gate into a Bayesian network rule.

Figure 5 is a diagram of sequential correlation gates first converted into priority AND gates and then into Bayesian network rules.

Figure 6 is a diagram showing the transformation of functional correlation gates into Bayesian network rules.

Fig. 7 is the Bayesian network model converted from the dynamic fault tree of the hydraulic end of the drilling pump.

Detailed ways

First, the present invention will be further described in conjunction with examples and accompanying drawings:

Step 1: Use the deductive reasoning method to establish the dynamic fault tree of the hydraulic end of the drilling pump:

1. Take the hydraulic end system failure of the drilling pump as the top event of the dynamic fault tree of the hydraulic end of the drilling pump, take the hydraulic end failure and power end failure of the drilling pump as the second-level intermediate events M1 and M2, and take each The faults that are prone to occur in each sub-component in the main component are regarded as the next-level event, and the faults of the components are analogized as the bottom event, and the top event, each intermediate event and the bottom event are coded, as shown in Table 1.

Table 1 Correspondence between event code and event name

event code event event code event event code event T Liquid end failure X1 The suction line is not tightly sealed X9 Cylinder head not tightened A Suction pipe failure X2 Clogged suction filter X10 Cylinder head cover loose B valve failure X3 Serious valve wear X11 Clogged discharge filter cartridge C Piston failure X4 The valve cover is not tightened X12 Blocked discharge line D Cylinder failure X5 Guide bush stuck X13 Inflatable connector blocked E Discharge pipe failure X6 Severely worn piston X14 ruptured air bag F Air bag failure X7 Piston nut loose X15 The globe valve is not tightly sealed G Safety valve failure X8 Cylinder enters air X16 Improperly set safety valve

Among them, T is the top event, A, B, C, D, E, F, G are the middle events, and X1-X15 are the bottom events.

2. According to the working principle and failure mechanism of the hydraulic end of the drilling pump, use the corresponding dynamic logic gate to connect the events at all levels with the top event, from the bottom event to the top event, use the dynamic logic gate to connect the events at all levels to obtain the drilling pump fluid. The dynamic fault tree of the force end is shown in Figure 2.

Step 2: According to the failure distribution model of the system, establish the corresponding relationship of "rate parameter λ--division number n"

1. First, transform the topology of the dynamic fault tree model into the network structure of the Bayesian network model. Then, according to the conditional probability table construction method in formula 1-6, use Matlab to compile the conditional probability table construction function of various logic gates, and then use the BNT toolbox and the variable elimination algorithm to solve the BN model.

2. Considering the problem of calculation amount, set the value range of n within [1,10], select different rate parameters λ to calculate the reliability of the system, and compare the maximum difference ratio of the system reliability between each group of data , the most suitable time division number n that should be selected in different ranges of the rate parameter λ can be obtained, and the corresponding relationship of "rate parameter λ--division number n" can be established.

Taking the rate parameter of λ∈(0.1×10 ^-6 h ^-1 , 4.5×10 ^-6 h ^-1 ) as an example, assuming that the task time t=50000h, when n is 2, 3, and 4, the system reliability varies with time The change curve is shown in Figure 3. It is obtained from the analysis of the calculated data: at each time point, the maximum difference between the data obtained by n is 3 and the data obtained by n is 2 is 0.0521%, the data obtained by n is 4 and the data obtained by n is 3 The maximum difference ratio of reliability is 0.0297%, so the time division number n corresponding to the rate parameter λ in the interval of (0.1×10 ^-6 h ^-1 , 4.5×10 ^-6 h ^-1 ) is 4.

Step 3: Define the state of the nodes in the network

Divide the time interval [0, t] composed of the entire task into n sub-intervals of equal length, and the length of each sub-interval is Δ=t/n, then the entire time axis [0, +∞) is divided into n+1 sub-intervals interval. When a component corresponding to a node A fails within the i-th time interval within the task time t, that is, A fails within the time interval [(i-1)Δ, iΔ], then the node A is said to be in state i. If A does not fail within task time t, that is, A fails within [t, ∞), then A is said to be in state n+1.

Through the above definition, the state space of all nodes in the Bayesian network is obtained as the following time interval:

[0,Δ],(Δ,2Δ],…,((n-1)Δ,nΔ],(nΔ,+∞). Abbreviated as {1,2,…,n+1}, the system and the invisible The failure time X always corresponds to a certain interval i of n+1 intervals. The sum of the probabilities of the system in the first n states is the failure rate of the system at the task time t, and the system is in the n+1th state. The probability is the non-failure rate of the system at task time t.

Step 4: Convert the dynamic fault tree model to a Bayesian network model;

According to the rules that the top event, middle event and bottom event of the dynamic fault tree correspond to the root node, middle node and leaf node of the Bayesian network model respectively, the dynamic fault tree event is transformed into a Bayesian network node, and the AND gate (AND), OR gate (OR), priority AND gate (PAND) are converted into Bayesian network model rules as shown in Figure 4, sequence correlation gate (SEQ) is first converted into priority AND gate and then converted into Bayesian network model as shown in Figure 5, The functional correlation gate (FDEP) was transformed into a Bayesian network model as shown in Figure 6. According to the above rules, the dynamic fault tree of the hydraulic end of the drilling pump is converted into a Bayesian network model as shown in Figure 7. In the Bayesian network model, there is a marginal probability distribution table (Marginal Probability Distribution, MPD) next to each root node, which lists all the states of the node and their corresponding probabilities. Each non-root node has a conditional probability distribution table (Conditional Probability Distribution, CPD), which records the conditional probability distribution of the node given the state combination of its parent nodes.

Step 5: Build the probability distribution of all nodes

That is, the quantitative analysis of dynamic fault tree fault diagnosis based on Bayesian network is completed.

AND gate

Let X=[X ₁ ,X ₂ ,...,X _m ], where m is the number of input events of the AND gate, X _i ,i=1,2,...,m is the state variable of the input event, and the number of state combinations thereof is (n+1) ^m , where n is the number of time divisions. Let Y be the state variable output by the AND gate, the state space of all variables is {1,2,...,n+1}, let k=max(X ₁ ,X ₂ ,...,X _m ). The failure mechanism of the AND gate is that the output event occurs when all input events occur, and the output event should be at the maximum value of the state values of all events. Therefore, in any state combination of input events, the conditional probability distribution of Y is:

Indicates that in the CPD table of the output node of the AND gate, the value of the element is 0 or 1, and only the element in the column corresponding to the maximum value of the input event state in each row is 1, and the other elements are 0.

OR gate

Similar to the AND gate, the difference is that the failure mechanism of the OR gate is that as long as one of the input events occurs, the output event occurs, so the state of the OR gate output event is the same as the minimum value of the state in the input event. Let r=min(X ₁ , X ₂ ,...,X _m ), then the conditional probability distribution of the OR gate output event is

The distribution is the same as the form of the AND gate, the difference is that the element in the column corresponding to the minimum value of the input event state in each row is 1, and the other elements are 0.

priority AND gate

Suppose the input events are A, B, the output event is Y, and the state values are a, b, and y, respectively. The failure mechanism of the priority AND gate is that when both A and B fail, A fails before B before event Y occurs, that is, when a < b, output event Y is in state b; otherwise, output event Y is in state n+1, which means that Y has not failed. The conditional probability distribution of Y is as follows:

When a < b,

When a≥b,

Equation 3 shows that when the input event satisfies the priority failure condition, the probability that the output event is in the state of the input event B is 1. Equation 4 indicates that when the input event does not satisfy the priority failure condition, the probability that the output event Y is in state n+1 is 1.

Sequence dependent gates:

Sequence-dependent gates force their input events to occur in a particular order and not fail in any other order. Sequential correlation gates are similar to priority AND gates in that they both represent the timing of basic events. The difference between them is that input events in sequential correlation gates cannot fail in any order; while priority AND gates can fail in any order, only in a specific order. will trigger the failure of its output event. Therefore, the sequential correlation gate is converted into a cascaded form of multiple priority AND gates during calculation.

function related gate

Assume that the function-related gate has only one related input event A, the trigger event is Tr, and the output event is T. When Tr occurs, its related event A occurs, and output event T occurs; when A fails independently, output event T also occurs. That is to say, whether event A is an independent failure or a related failure, it will lead to the occurrence of output event T. In order to determine the CPD table (conditional probability distribution table) of the node more intuitively and conveniently, an intermediate node A' is added between the input event and the output event to represent the failure of A caused by Tr triggering or the independent failure of A itself. Total failure events (if this node is not added, the failure probability distribution of node A will no longer be a marginal distribution, but a conditional probability distribution affected by Tr). Therefore, its conditional probability distribution table is a unit matrix E, and its conditional probability distribution is:

Step 6: Reverse reasoning to obtain fault diagnosis results

According to the conditional independence of the Bayesian network, the conditional probability distribution of each node can be expressed as P(X _i |pa(X _i )) to express the quantitative relationship between the node and the parent node. Given the prior probability distribution of the root node and the conditional probability distribution of the non-root nodes, the joint probability distribution including all nodes can be obtained, and then the marginalized calculation of the target node is performed to obtain its marginal probability distribution. Finally, using the reverse reasoning of the Bayesian network, the failure probability of each bottom event is calculated, and the diagnosis results are output according to the size of the failure probability.

In a DTBN containing m nodes, non-leaf nodes are represented by events U _i (1≤i≤m-1), and the occurrence interval of U _i is {[0,Δ],(Δ,2Δ],...,( (n-1)Δ,nΔ],(T,+∞)}.If the top event U _T occurs within the task time T, then the top event must occur at [0,Δ],(Δ,2Δ], ...,((n-1)Δ,nΔ],(nΔ,+∞) in one of the intervals. Therefore, the probability of U _T occurring within the task time T can be directly calculated and expressed as:

In the formula, _ui represents the occurrence interval of U _i , and _ui belongs to {[0,Δ],(Δ,2Δ],...,((n-1)Δ,nΔ],(T,+∞)} .

Then the failure probability of each component can be obtained when the system fails by using the method of reverse reasoning.

The rate parameter λ of the exponential distribution approximated by the hydraulic end fault distribution function of the drilling pump is within the interval (0.1×10 ^-6 h ^-1 , 4.5×10 ^-6 h ^-1 ), so the time division number n is taken as 4. When the system fails, that is, when the state of the leaf node is 5, the reverse reasoning of the Bayesian network is used to calculate the failure probability of each bottom event as shown in Table 2.

Table 2 Failure probability of each bottom event when the top event occurs

event code Failure probability event code Failure probability X1 0.0231 X9 0.0561 X2 0.0261 X10 0.0472 X3 0.1090 X11 0.0842 X4 0.0354 X12 0.0233 X5 0.0261 X13 0.0463 X6 0.0116 X14 0.0321 X7 0.1562 X15 0.0569 X8 0.1008 X16 0.0103

From this table, the probability of failure of X16 is the smallest, while the probability of failure of X7, X3, and X8 is the largest. Therefore, the failure probability of loose piston nut, serious valve wear, and air entering the cylinder is more likely.

Claims (3)

1. A method for diagnosing faults at the hydraulic end of a drilling pump based on a dynamic fault tree, the method comprising: Step 1: Use the deductive reasoning method to establish the dynamic fault tree of the drilling pump hydraulic system: Step 1.1: Take the failure of the hydraulic end of the drilling pump as the top event of the dynamic fault tree, take the failure of the sub-components in the main component in the hydraulic end as the intermediate event, and the failure of the components in the sub-component as the bottom event; Step 1.2: According to the working principle and failure mechanism of the hydraulic end of the drilling pump, use the corresponding dynamic logic gate to connect the events at all levels with the top event, from the bottom event to the top event, use the dynamic logic gate to connect the events at all levels to obtain the drilling pump. Dynamic fault tree of the liquid end; Step 2: Convert the dynamic fault tree into a Bayesian network model, and establish the corresponding relationship of "rate parameter λ--time division number n"; Step 2.1: Convert the dynamic fault tree into a Bayesian network model; the top event, middle event, and bottom event of the dynamic fault tree correspond to the root node, middle node, and leaf node of the Bayesian network model, respectively. The expression method of logic gate in Bayesian network model is: AND gate: Let X=[X ₁ ,X ₂ ,...,X _m ], where m is the number of input events of the AND gate, X _i ,i=1,2,...,m is the state variable of the input event, and the number of state combinations thereof is (n +1) ^m ; let Y be the state variable output by the AND gate, the state space of all variables is {1,2,...,n +1}, let k=max(X ₁ ,X ₂ ,..., X _m ), j is the judgment value; the failure mechanism of the AND gate is that the output event occurs when all input events occur, then the output event should be at the maximum value of the state values of all events, therefore, in any state combination of input events, Y The conditional probability distribution of is:

OR gate: The failure mechanism of the OR gate is that as long as one of the input events occurs, the output event occurs, so the state of the OR gate output event is the same as the minimum value of the state in the input event; let r=min(X ₁ ,X ₂ ,...,X _m ), then the conditional probability distribution of the OR gate output event is:

Priority AND gate: Assuming that the input events are A and B, the output event is Y, and the state values are a, b, and y, respectively, the failure mechanism of the priority AND gate is that when both A and B fail, A fails before B, and event Y will fail. occurs, that is, when a<b, the output event Y is in state b; otherwise, the output event Y is in state n + 1, which means that Y has not failed; then the conditional probability distribution of the output event Y of the priority AND gate is as follows: When a < b,

When a≥b,

Sequence dependent gates: The sequential correlation gate forces its input events to occur in a specific order, and will not fail in other orders; the sequential correlation gate is similar to the priority AND gate, both of which represent the timing of basic events. The difference between them is: the sequential correlation gate The input events cannot fail in any order; the priority AND gate can fail in any order, and only the failure of a specific order will trigger the failure of its output event; therefore, the sequence-dependent gate can be changed to a cascade of multiple priority AND gates; Function related gates: Suppose the function-related gate has only one related input event A, the trigger event is Tr, and the output event is T; when Tr occurs, its related event A occurs, and output event T occurs; when A fails independently, output event T also occurs; An intermediate node A' is added between the input event and the output event to represent the total failure event of A failure caused by Tr triggering or the independent failure of A itself, p, q are the corresponding event states, then the function-related gate output The conditional probability distribution of the event is:

Step 2.2: Set the value range of n, use multiple sets of data collected in advance, and select different rate parameters λ to calculate the reliability of the system; compare the maximum difference ratio of the system reliability between each group of data, the rate parameter can be obtained λ should select the most suitable time division number n in different ranges, and establish the corresponding relationship of "rate parameter λ--division number n"; Step 3: Define the state of the nodes in the Bayesian network model; Divide the time interval [0, t] composed of the entire task into n sub-intervals of equal length, and the length of each sub-interval is Δ=t/n, then the entire time axis [0, +∞) is divided into n+1 sub-intervals interval; when a component corresponding to a node A fails within the i-th time interval within the task time t, that is, A fails within the time interval [(i-1)Δ, iΔ], then node A is said to be in state i ; If A does not fail within task time t, that is, A fails within [t, ∞), then A is in state n+1; Through the above definition, the state space of all nodes in the Bayesian network is obtained as the following time interval: [0,Δ],(Δ,2Δ],…,((n-1)Δ,nΔ],(nΔ,+∞), abbreviated as {1,2,...,n+1}, the system And the failure time X of the component always corresponds to a certain interval i of n+1 intervals; the sum of the probabilities that the system is in the first n states is the failure rate of the system at the task time t, and the system is in the n+1th The probability of the state is the non-failure rate of the system at task time t; Step 4: establish the probability distribution of all nodes according to the state of the nodes in the Bayesian network model, and complete the quantitative calculation of the dynamic fault tree fault diagnosis based on the Bayesian network; Step 5: Reverse reasoning to obtain fault diagnosis results; The conditional probability distribution of each node is expressed as P(X _i |pa(X _i )), which is used to express the quantitative relationship between child nodes and parent nodes; the prior probability distribution at a given root node and the non-root node’s Under the condition of conditional probability distribution, the joint probability distribution including all nodes is obtained, and then the target node is marginalized to obtain its marginal probability distribution; finally, the failure probability of each bottom event is calculated by the reverse reasoning of Bayesian network. , arrange according to the size of the failure probability, and output the diagnosis results. 2. The method for diagnosing the hydraulic end of a drilling pump based on a dynamic fault tree as claimed in claim 1, wherein the top event in the step 1.1 is a hydraulic end failure, and the middle event is: a suction pipe failure , valve failure, piston failure, cylinder failure, discharge pipe failure, air bag failure, safety valve failure, the bottom events are: the suction pipeline is not tightly sealed, the suction filter is blocked, the valve is seriously worn, the valve cover is not tightened, The guide sleeve is stuck, the piston is seriously worn, the piston nut is loose, the cylinder enters the air, the cylinder head is not tightened, the cylinder head gland is loose, the discharge filter cartridge is blocked, the discharge pipeline is blocked, the inflation joint is blocked, and the inner bag of the air bag is ruptured , The shut-off valve is not tightly sealed, and the safety valve is improperly set; The relationship between the events in step 1.2 is: The suction pipeline is not tightly sealed, the suction filter screen is blocked and the door is connected to the suction pipe fault; the valve is seriously worn, the valve cover is not tightened, the guide sleeve is stuck through or the door is connected to the valve fault; the piston is seriously worn, and the piston nut Looseness is connected to the fault of the piston through the OR door; the cylinder enters the air, the cylinder head is not tightened, the cylinder head gland is loose and is connected to the fault of the cylinder through the sequence-related door; the discharge filter cartridge is blocked, the discharge line is blocked, and the or door is connected to the discharge pipe The fault is connected; the inflation joint is blocked, the inner bag of the air bag is ruptured, and the shut-off valve is not tightly sealed. It is connected to the air bag fault through the function-related door. The rupture of the inner bag is the trigger event input of the function-related door, and the air bag failure is the output of the function-related door; the improper setting of the safety valve is connected with the door and the safety valve fault. 3. a kind of fault diagnosis method of drilling pump hydraulic end based on dynamic fault tree as claimed in claim 1, is characterized in that, the concrete method of described step 5 is: In the discrete-time Bayesian network DTBN containing m nodes, the non-leaf nodes are represented by events U _i , where 1≤i≤m-1, and the occurrence interval of U _i is {[0,Δ],(Δ,2Δ] ,...,((n-1)Δ,nΔ],(T,+∞)}; if the top event U _T occurs within the task time T, then the top event must occur at [0,Δ], (Δ,2Δ],...,((n-1)Δ,nΔ],(nΔ,+∞) in one of the intervals; therefore, the probability of U _T occurring within the task time T can be directly calculated and expressed for:

In the formula, _ui represents the occurrence interval of U _i , and _ui belongs to {[0,Δ],(Δ,2Δ],...,((n-1)Δ,nΔ],(T,+∞)} ; Then the failure probability of each component can be obtained when the system fails by using the method of reverse reasoning.

Images