CN115452391A - Diesel engine exhaust temperature abnormity diagnosis method based on neural network expert system - Google Patents

Abstract

The invention aims to provide a method for diagnosing the exhaust temperature abnormity of a diesel engine based on a neural network expert system, which comprises the following steps: acquiring related operation data aiming at the problem of abnormal exhaust temperature of the diesel engine, and dividing a data set into two parts, wherein one part is used as training data, and the other part is used as test data; describing both training data and test data as a knowledge system; normalizing the fault feature matrix to obtain a normalized fault feature matrix; establishing a BP neural network; establishing an expert knowledge base according to an output result of the BP neural network; and verifying the model precision. The invention can accurately diagnose the exhaust temperature abnormity of the diesel engine; the problems that the expert system knowledge acquisition period is long and the neural network reasoning process and the result are difficult to understand are solved; the invention adopts a particle swarm algorithm to optimize two key parameters in a BP neural network model: and connecting the weight value with an output threshold value, so that the classification precision and the calculation time of the model are improved.

Description

Diagnosis Method of Diesel Engine Exhaust Temperature Abnormality Based on Neural Network Expert System

technical field

The invention relates to a diesel engine fault diagnosis method, in particular to a diesel engine exhaust temperature diagnosis method.

Background technique

Diesel engine is currently the most widely used model among all kinds of power machinery. The development of diesel engine technology has an important impact on my country's industry and agriculture, transportation, national defense construction and other aspects. The abnormal exhaust temperature of the diesel engine is a specific symptom of the internal failure of the diesel engine. It may involve single system single fault, single system multiple faults and multiple system multiple faults. It has a complex coupling relationship. If it is not diagnosed and repaired in time, it will lead to mechanical efficiency of the diesel engine The reduction in power will affect the safe operation of equipment and systems, and even affect the personal safety of operators. Therefore, it is of great significance to carry out diagnostic research on abnormal exhaust temperature to ensure the safe and efficient operation of diesel engines.

The document "Example of Abnormal Exhaust Temperature of Wartsila Auxiliary Engine" ("Navigation Technology", 2020) diagnoses abnormal exhaust temperature of Wartsila diesel engine, checks parameters such as scavenging temperature and turbine speed, disassembles and checks speed regulation Check the relevant parts between the gear and the rack of the high-pressure oil pump, and check the maintenance record book, and finally determine that the fault is the reduction of the cylinder throttle. The diagnosis method used in this document is in the stage of post-event maintenance. Its disadvantages are long diagnosis cycle, cumbersome steps, backward diagnosis methods, and may destroy the best working state of the diesel engine during the inspection and maintenance process. The document "MTU396 Diesel Engine Exhaust Temperature Difference Analysis and Troubleshooting" ("Internal Combustion Engine", 2021) aimed at the problem of large exhaust temperature difference between A and B columns, disassembled the intake pipe cover, exhaust connection pipe, A1~A4 and B1 The 8 fuel injectors and other related parts of ~B4 still have faults after adjusting the fuel injection advance angle and resetting the relevant parts. After disassembling the fuel injection pump and fuel injector test again, it is determined that the fault point is the oil return line check valve The spool is stuck. The diagnostic method used in this document repeatedly disassembled the fault-related components, and carried out many experiments in the presence of faults. The damage to the diesel engine was serious, and the troubleshooting process was complicated and cumbersome, and the technical means were backward.

Contents of the invention

The object of the present invention is to provide a diesel engine exhaust temperature abnormal diagnosis method based on a neural network expert system that can overcome the shortcomings of the long expert system knowledge acquisition period and the difficulty in understanding the neural network reasoning process and results during the diagnosis process.

The purpose of the present invention is achieved like this:

The present invention is based on the abnormal diagnosis method of diesel engine exhaust temperature of neural network expert system, is characterized in that:

(1) Aiming at the problem of abnormal diesel engine exhaust temperature, obtain relevant operating data, and divide the data set into two parts, one of which is used as training data T _r , and the other is used as test data T _e ;

(2) Describe the training data T _r and test data T _e in step (1) as a knowledge system ∑=(F, S), where the output result set F={f ₁ , f ₂ ,..., f _n } is the typical state of the diesel engine, the input parameter set S={s ₁ , s ₂ ,…, s _k } is the parameters in the actual operation of the diesel engine, and the training fault characteristic matrix R _r (∑)=(r _ij ) is established according to the training data T _{r n×k} in preparation for training the neural network; establish a test fault feature matrix _Re (∑)=(r _ij ) _n×k according to the test data T _e in preparation for testing the neural network expert system after it is established, Where r _ij is the value of the relevant parameter s _i in the actual operation of the diesel engine, the value range of i is {i∈Z|1≤i≤n}, and the value range of j is {j∈Z|1≤j≤k} ;

(3) Normalize the fault characteristic matrix R _r (∑) and _Re (∑) in step (2), and obtain the normalized fault characteristic matrix R′ _r (∑)=(r′ _ij ) _n×k and R′ _e (∑)=(r′ _ij ) _n×k , where r′ _ij is the normalized value of the relevant parameter s _i in the actual operation of the diesel engine, and its value range is [0, 1 ], establish training examples and test examples according to the fault feature matrix;

与输出阈值θ_m、

的寻优过程，得到最优连接权值ρ_jm、

与输出阈值θ_m、

输出BP神经网络模型；(4) Establish a BP neural network and use the training examples in step (3) to train the BP neural network, and use the particle swarm optimization algorithm to accelerate the connection weights ρ _jm ,

and output threshold θ _m ,

In the optimization process, the optimal connection weights ρ _jm ,

and output threshold θ _m ,

Output BP neural network model;

(5) Based on the BP neural network obtained in step (4), the connection between it and the expert system is established, that is, the expert knowledge base is established with the output result of the BP neural network, and the structure of the inference engine and the interpreter expert system are established, and the neural network is output. Network expert system model;

(6) Use the neural network expert system model obtained in step (5) and the test case in step (3) to test and verify the accuracy of the model.

The present invention may also include:

1. Obtaining relevant operating data in step (1) includes obtaining operating data of 7 groups of related measuring points, which are the data when the diesel engine is in normal operation and the operating data when there is one of the following faults: the fault type is supercharging Injector damage fault, fuel injection abnormality, intake pipe blockage fault, exhaust valve damage fault, fresh water system abnormality and high-pressure fuel pipe leakage fault.

2. The output result set F in step (2) includes: normal operation state f ₁ and fault state, including supercharger damage f ₂ , fuel injection abnormality f ₃ , intake pipe blockage f ₄ , exhaust valve damage f ₅ , Fresh water system abnormality f ₆ , high-pressure oil pipe leakage f ₇ , that is, F={f ₁ , f ₂ , f ₃ , f ₄ , f ₅ , f ₆ , f ₇ };

The input parameter set S includes: exhaust gas temperature s ₁ , turbocharger speed s ₂ , fuel injection status signal s ₃ , air cooler outlet air pressure s ₄ , exhaust valve status signal s ₅ , fresh water inlet pressure s ₆ , high pressure fuel oil Oil leakage level s ₇ , that is, S={s ₁ , s ₂ , s ₃ , s ₄ , s ₅ , s ₆ , s ₇ };

The forms of the fault characteristic matrix R _r (∑) and _Re (∑) are both:

The rows represent seven typical states that may occur in the diesel engine system, and the columns represent seven relevant operating parameters of the diesel engine that can be monitored.

3. In step (3), normalize the fault characteristic matrix R _r (∑) and _Re (∑), that is, use the maximum and minimum values for each column of the fault characteristic matrix R _r (∑) and _Re (∑) The method is normalized:

Among them, i (i=1, ..., 7) is the state type identification, j (j = 1, ..., 7) is the parameter type identification, the value r' _ij after normalization is in [0, 1], the fuel injection state Signal s ₃ and exhaust valve status signal s ₅ are 1 in normal state and 0 in abnormal state;

Establish a training example or a test example, that is, take a row of the normalized fault feature matrix as the first dimension of the training example or test example, that is, the input vector, set the state corresponding to the row to 1, and set the other states to 0 , constituting the second dimension of the training example or test example, that is, the expected output vector, a total of 7 training examples are established; taking the intake pipe blockage fault f ₄ as an example, the forms of the training example and the test example are as follows:

(x ₄ , y ₄ )=((r′ ₄₁ , r′ ₄₂ , r′ ₄₃ , r′ ₄₄ , r′ ₄₅ , r′ ₄₆ , r′ ₄₇ ) ^T , (0,0,0,1,0 , 0, 0) ^T ).

4, use particle swarm optimization algorithm to optimize BP neural network in step (4) and comprise the following steps:

(a) Initialize the BP neural network model:

Using a neural network structure with one hidden layer, the number k of neurons in the input layer, the number n of neurons in the output layer, and the number q of neurons in the hidden layer are determined by the following formula:

Among them, t is a constant, t∈[0,10];

Set the input of the mth (1≤m≤q) neuron node of the hidden layer as:

Among them, _ρjm is the connection weight between the jth neuron node in the input layer and the mth neuron node in the hidden layer, and r′ _ij is the normalized value of the jth input parameter in the state i;

Set the output of the mth neuron node in the hidden layer as:

为隐含层第m个神经元节点的输出阈值；in

is the output threshold of the mth neuron node in the hidden layer;

Set the i-th neuron node of the output layer, that is, the input of the i-th sub-vector in the output vector is:

为隐含层第m个神经元节点与输出层第i个神经元节点的连接权重，b_h为第m个隐含层神经元节点的输出；in

is the connection weight between the mth neuron node of the hidden layer and the ith neuron node of the output layer, b _h is the output of the mth hidden layer neuron node;

Set the i-th neuron node of the output layer, that is, the value of the i-th sub-vector in the output vector is:

Where θ _i is the output threshold of the i-th neuron node in the output layer;

(b) Initialize the particle swarm.

Set two particle swarm stop iteration conditions, the first is the upper limit of the number of iterations, and the second is that the confidence of the BP neural network reaches 95%:

The formula for setting the speed and position update of each particle is:

为第k+1代粒子u的速度在第d维的上的分量，

为第k代粒子u的个体极值在第d维的分量，

为第k代粒子的全局极值在第d维的分量，

为第k代粒子u的位置在第d维的上的分量，c₁与c₂均为学习因子，ω为惯性权重，ζ和η为(0，1)上随机数；in

is the component of the velocity of the k+1th generation particle u on the d-th dimension,

is the d-dimensional component of the individual extremum of particle u of the kth generation,

is the component of the d-th dimension of the global extremum of the k-th generation particle,

is the component of the position of the kth generation particle u _on the d dimension, c1 and c2 are learning factors, ω is the inertia weight, and ζ and η are random numbers on (0, ₁ );

Set the inertia weights to:

k is the number of iterations, k _max is the maximum number of iterations, in the initial state ω=ω _max =1, ω _min =0.4;

The fitness function of the particles is set as:

表示该粒子在第i个训练例下的输出结果，y_i为预期输出结果；Among them, X is the position of a particle in the particle swarm, n is the number of training examples,

Indicates the output result of the particle under the i-th training example, and y _i is the expected output result;

与输出阈值θ_m、

作为粒子群中每个粒子的位置分量，粒子群中第u个粒子位置为：The particle velocity is initialized, and the value is a uniformly distributed random number between [v _min , v _max ], which is [-0.05, 0.05] in the present invention. Take the connection weight ρ _ij in the BP neural network,

and output threshold θ _m ,

As the position component of each particle in the particle swarm, the position of the uth particle in the particle swarm is:

Among them, x _us is the sth position component of the uth particle, and the value range is [x _min , x _max ]. When a certain position component exceeds the upper or lower bound value, this component is approximately taken as the boundary value, The value range of s is {s∈Z|1≤s≤q(k+n+1)+n}, that is, the value range of s is {s∈Z|1≤s≤67};

(c) If the current fitness function value of the particle is better than the individual historical optimal value, update the current position of the particle to the individual historical optimal position; if the individual historical optimal position of the particle is better than the global optimal position, then Update the global optimum of the particle to the current individual historical optimum;

(d) Update the position and velocity of each particle according to the particle swarm optimization algorithm;

(e) If any one of the stop iteration conditions is met, stop the iteration, get the optimal initial weight and threshold of the BP network, and output the trained BP neural network.

5, in the step (5), based on the BP neural network model, the method for establishing its connection with the expert system is:

Investigate the hidden layer neurons pointing to the output of the BP neural network, and find out the positive weighted input value of the output layer, which has the following characteristics:

At the same time, examine the input layer neurons pointing to this hidden layer neuron, and also find out its positive weighted input value to form a rule:

IF expression 1 AND expression 2 AND...AND expression n THEN conclusion

The specific form of the expression n in the rule is that the output of a neuron in the input layer is positive, that is, there is a certain symptom, and the specific form of the conclusion in the rule is the fault type; the following rules are obtained for the abnormal temperature of the exhaust gas of the diesel engine:

IF exhaust temperature > 550°C AND supercharger speed reduction THEN supercharger damage

IF exhaust temperature > 550°C AND air cooler outlet pressure decreases THEN intake pipe is clogged

IF exhaust temperature > 550°C AND abnormal fuel injection sign = 1 THEN abnormal fuel injection

IF Exhaust temperature > 550°C AND Exhaust valve damaged flag = 1 THEN Exhaust valve damaged

IF exhaust temperature > 550°C AND fresh water inlet pressure decreases THEN fresh water system is abnormal

IF exhaust temperature is low AND high pressure fuel pipe leaks oil level rises THEN high pressure fuel pipe leaks

Store such rules in the knowledge base to form the knowledge base of the neural network expert system, and build an expert system based on this knowledge base.

6. The process of testing the neural network expert system in step (6) is:

(i) Receive the alarm signal of abnormal exhaust temperature of the diesel engine;

(ii) The inference engine of the expert system refers to the knowledge base, performs fault diagnosis for the abnormal data of the associated measurement points, and calls the neural network module;

(iii) If the confidence degree of the diagnosis result of the neural network is greater than 0.95, it will be regarded as the final diagnosis result of the neural network expert system, and the diagnosis result will be explained by the expert system;

(iv) If the output confidence of the neural network is not greater than 0.95, then output the diagnosis result of the expert system;

(v) If neither the expert system nor the neural network can diagnose the fault, that is, there is a fault that cannot establish the corresponding relationship between symptoms and causes at this stage, then pack the data, start the learning mode of the neural network expert system, and let the expert name the fault , adjust the neural network topology, increase the output nodes, and use the packaged offline data to train the neural network, and update the trained neural network to the neural network expert system.

The advantages of the present invention are:

1. The present invention can accurately diagnose the abnormal temperature of the exhaust gas of the diesel engine;

2. Solved the problems of long acquisition cycle of expert system knowledge and difficult understanding of neural network reasoning process and results;

与输出阈值θ_m、

这提高了模型的分类精度和计算时间。3. The present invention optimizes two key parameters in the BP neural network model by applying the particle swarm optimization algorithm: connection weight ρ _jm ,

and output threshold θ _m ,

This improves the classification accuracy and computation time of the model.

Description of drawings

Fig. 1 is a structural representation of the present invention;

Fig. 2 is a schematic flow chart of the present invention.

detailed description

The present invention is described in more detail below in conjunction with accompanying drawing example:

In conjunction with Fig. 1-2, a kind of neural network expert system fault diagnosis method of the present invention for abnormal exhaust temperature of diesel engine comprises the following steps:

Step 1: Aiming at the problem of abnormal diesel engine exhaust temperature, obtain relevant operating data, and divide the data set into two parts, one of which is used as training data T _r , and the other is used as test data T _e ;

Step 2: Describe the training data T _r and test data T _e in step 1 as a knowledge system ∑=(F, S), where the output result set F={f ₁ , f ₂ ,..., f _n } is the diesel engine In a typical state, the input parameter set S={s ₁ , s ₂ ,…, s _k } is the relevant parameters in the actual operation of the diesel engine, and the training fault feature matrix R _r (∑)=(r _ij ) _n is established according to the training data T _{r ×k} for training the neural network; establish the test fault feature matrix _Re (∑)=(r _ij ) _n×k according to the test data T _e for testing the neural network expert system after it is established, where r _ij is the value of the relevant parameter s _i in the actual operation of the diesel engine, the value range of i is {i∈Z|1≤i≤n}, and the value range of j is {j∈Z|1≤j≤k};

Step 3: Normalize the fault characteristic matrix R _r (∑) and _Re (∑) in step 2, and obtain the normalized fault characteristic matrix R′ _r (∑)=(r′ _ij ) _{n ×k} and R′ _e (∑)=(r′ _ij ) _n×k , where r′ _ij is the normalized value of the relevant parameter s _i in the actual operation of the diesel engine, and its value range is [0, 1], Establish training examples and test examples according to the fault feature matrix;

与输出阈值θ_m、

的寻优过程，得到最优连接权值ρ_jm、

与输出阈值θ_m、

输出BP神经网络模型；Step 4: Establish a BP neural network and use the training examples in step 3 to train the BP neural network, and use the particle swarm optimization algorithm to accelerate the connection weights ρ _jm ,

and output threshold θ _m ,

In the optimization process, the optimal connection weights ρ _jm ,

and output threshold θ _m ,

Output BP neural network model;

Step 5: Based on the BP neural network obtained in step 4, establish the connection between it and the expert system, that is, establish an expert knowledge base with the output of the BP neural network, and establish an expert system structure such as an inference engine and an interpreter, and output the neural network Expert system model;

Step 6: Use the neural network expert system model obtained in step 5 and the test case in step 3 to test and verify the accuracy of the model.

To establish a data set in step 1, it is necessary to obtain 7 sets of operating data of related measuring points, which are the data when the diesel engine is in normal operation and the operating data when there is one of the following faults. The fault types are supercharger damage fault, injection Abnormal oil, intake pipe blockage, exhaust valve damage, fresh water system abnormality and high-pressure oil pipe leakage;

The output result set F in step 2 includes: normal operation state f ₁ , and fault states, including supercharger damage f ₂ , fuel injection abnormality f ₃ , intake pipe blockage f ₄ , exhaust valve damage f ₅ , fresh water system abnormality f ₆ , high-pressure oil pipe leakage f ₇ , that is, F={f ₁ , f ₂ , f ₃ , f ₄ , f ₅ , f ₆ , f ₇ };

The input parameter set S includes: exhaust gas temperature s ₁ , turbocharger speed s ₂ , fuel injection status signal s ₃ , air cooler outlet air pressure s ₄ , exhaust valve status signal s ₅ , fresh water inlet pressure s ₆ , high pressure fuel oil Pipe oil leakage level s ₇ , that is, S={s ₁ , s ₂ , s ₃ , s ₄ , s ₅ , s ₆ , s ₇ }.

The forms of the fault characteristic matrix R _r (∑) and _Re (∑) are both:

The rows represent seven typical states that may occur in the diesel engine system, and the columns represent seven relevant operating parameters of the diesel engine that can be monitored.

Multi-parameter selection is beneficial to comprehensively judge the operation status of the diesel engine and make the diagnosis results more accurate. Due to the diversification of reference parameters, the fault diagnosis model can also comprehensively analyze the multi-parameter information of the diesel engine and improve the diagnostic accuracy and reliability.

In step 3, the fault characteristic matrix R _r (∑) and _Re (∑) are normalized, that is, each column of the fault characteristic matrix R _r (∑) and _Re (∑) is normalized by the maximum and minimum method Processing:

Among them, i (i=1, ..., 7) is the state type identification, j (j = 1, ..., 7) is the parameter type identification, the value r' _ij after normalization is in [0, 1], the fuel injection state The signal s ₃ and the exhaust valve status signal s ₅ are 1 in normal state and 0 in abnormal state. Normalization can eliminate the impact of dimension and order of magnitude differences between indicators.

Establish a training example or a test example, that is, take a row of the normalized fault feature matrix as the first dimension of the training example or test example, that is, the input vector, set the state corresponding to the row to 1, and set the other states to 0 , which constitutes the second dimension of the training or testing example, that is, the expected output vector, and a total of 7 training examples are established. Taking the intake pipe blockage fault _f4 as an example, the training examples and test examples are as follows:

(x ₄ , y ₄ )=((r′ ₄₁ , r′ ₄₂ , r′ ₄₃ , r′ ₄₄ , r′ ₄₅ , r′ ₄₆ , r′ ₄₇ ) ^T , (0,0,0,1,0 ,0,0) ^T )

Using the particle swarm optimization algorithm to optimize the BP neural network in step 4 includes the following steps:

(1) Initialize the BP neural network model.

The present invention adopts the neural network structure that contains one layer of hidden layer, input layer neuron number k, output layer neuron number n, hidden layer neuron number q is determined by following formula:

Wherein, t is a constant, t∈[0,10], in the present invention n=7, k=7, adjustment number t=0, so the neuron format q of the hidden layer is approximately taken as 4.

Set the input of the mth (1≤m≤q) neuron node of the hidden layer as:

Among them, _ρjm is the connection weight between the jth neuron node in the input layer and the mth neuron node in the hidden layer, and r′ _ij is the normalized value of the jth input parameter in the state i.

Set the output of the mth neuron node in the hidden layer as:

为隐含层第m个神经元节点的输出阈值。in

is the output threshold of the mth neuron node in the hidden layer.

Set the i-th neuron node of the output layer, that is, the input of the i-th sub-vector in the output vector is:

为隐含层第m个神经元节点与输出层第i个神经元节点的连接权重，b_h为第m个隐含层神经元节点的输出。in

is the connection weight between the mth neuron node of the hidden layer and the ith neuron node of the output layer, b _h is the output of the mth hidden layer neuron node.

Set the i-th neuron node of the output layer, that is, the value of the i-th sub-vector in the output vector is:

Where θ _i is the output threshold of the i-th neuron node in the output layer.

(2) Initialize the particle swarm.

Set two particle swarm stop iteration conditions. The first is the upper limit of the number of iterations, and the second is the 95% confidence level of the BP neural network.

The formula for setting the speed and position update of each particle is:

为第k+1代粒子u的速度在第d维的上的分量，

为第k代粒子u的个体极值在第d维的分量，

为第k代粒子的全局极值在第d维的分量，

为第k代粒子u的位置在第d维的上的分量，c₁与c₂均为学习因子，ω为惯性权重，ζ和η为(0，1)上随机数。in

is the component of the velocity of the k+1th generation particle u on the d-th dimension,

is the d-dimensional component of the individual extremum of particle u of the kth generation,

is the component of the d-th dimension of the global extremum of the k-th generation particle,

is the component of the position of the k-th generation particle u on the d-th dimension, c ₁ and c ₂ are learning factors, ω is the inertia weight, ζ and η are random numbers on (0, 1).

Set the inertia weights to:

k is the number of iterations, k _max is the maximum number of iterations, in the initial state ω=ω _max =1, ω _min =0.4.

The fitness function of the particles is set as:

表示该粒子在第i个训练例下的输出结果，y_i为预期输出结果。Among them, X is the position of a particle in the particle swarm, n is the number of training examples,

Indicates the output result of the particle under the i-th training example, and y _i is the expected output result.

与输出阈值θ_m、

作为粒子群中每个粒子的位置分量，粒子群中第u个粒子位置为：The particle velocity is initialized, and the value is a uniformly distributed random number between [v _min , v _max ], which is [-0.05, 0.05] in the present invention. Take the connection weight ρ _ij in the BP neural network,

and output threshold θ _m ,

As the position component of each particle in the particle swarm, the position of the uth particle in the particle swarm is:

Among them, x _us is the sth position component of the uth particle, and the value range is [x _min , x _max ], which is taken as [-1, 1] in the present invention, when a certain position component exceeds the upper bound or lower bound value , the component needs to be approximately taken as a boundary value, and the value range of s is {s∈Z|1≤s≤q(k+n+1)+n}, that is, the value range of s is {s∈Z |1≤s≤67}.

(3) If the current fitness function value of the particle is better than the individual historical optimal value, update the current position of the particle to the individual historical optimal position; if the individual historical optimal position of the particle is better than the global optimal position, then Update the particle's global optimum to the current individual historical optimum.

(4) Update the position and speed of each particle according to the particle swarm algorithm.

(5) If any stop iteration condition is satisfied, then stop the iteration, get the optimal initial weight and threshold of the BP network, and output the trained BP neural network.

In step 5, based on the BP neural network model, the method of establishing its connection with the expert system is:

Investigate the hidden layer neurons pointing to the output of the BP neural network, and find out the positive weighted input value of the output layer, which has the following characteristics:

At the same time, examine the input layer neurons pointing to this hidden layer neuron, and also find out its positive weighted input value to form a rule:

IF expression 1 AND expression 2 AND...AND expression n THEN conclusion

The specific form of the expression n in the rule is that the output of a neuron in the input layer is positive, that is, it has a certain symptom, and the specific form of the conclusion in the rule is the fault type. The following rules can be obtained for abnormal diesel exhaust temperature:

IF exhaust temperature > 550°C AND supercharger speed reduction THEN supercharger damage

IF exhaust temperature > 550°C AND air cooler outlet pressure decreases THEN intake pipe is clogged

IF exhaust temperature > 550°C AND abnormal fuel injection sign = 1 THEN abnormal fuel injection

IF Exhaust temperature > 550°C AND Exhaust valve damaged flag = 1 THEN Exhaust valve damaged

IF exhaust temperature > 550°C AND fresh water inlet pressure decreases THEN fresh water system is abnormal

IF exhaust temperature is low AND high pressure fuel pipe leaks oil level rises THEN high pressure fuel pipe leaks

Store such rules in the knowledge base to form the knowledge base of the neural network expert system, and build an expert system based on this knowledge base.

The process of testing the neural network expert system in step 6 is:

(1) Receive the alarm signal of abnormal exhaust temperature of the diesel engine;

(2) The inference engine of the expert system refers to the knowledge base, performs fault diagnosis for the abnormal data of the associated measurement points, and calls the neural network module;

(3) If the confidence degree of the diagnosis result of the neural network is greater than 0.95, it will be used as the final diagnosis result of the neural network expert system, and the diagnosis result will be explained by the expert system;

(4) If the output confidence of the neural network is not greater than 0.95, then output the diagnosis result of the expert system;

(5) If neither the expert system nor the neural network can diagnose the fault, that is, there is a fault that cannot establish the corresponding relationship between symptoms and causes at this stage, then pack the data, start the learning mode of the neural network expert system, and let the expert name the fault , adjust the neural network topology, increase the output nodes, and use the packaged offline data to train the neural network, and update the trained neural network to the neural network expert system.

Claims (7)

1. The abnormal diagnosis method of diesel engine exhaust temperature based on neural network expert system, is characterized in that: (1) Aiming at the problem of abnormal diesel engine exhaust temperature, obtain relevant operating data, and divide the data set into two parts, one of which is used as training data T _r , and the other is used as test data T _e ; (2) Describe the training data T _r and test data T _e in step (1) as a knowledge system ∑=(F, S), where the output result set F={f ₁ , f ₂ ,..., f _n } is the typical state of the diesel engine, the input parameter set S={s ₁ , s ₂ ,…, s _k } is the parameters in the actual operation of the diesel engine, and the training fault characteristic matrix R _r (∑)=(r _ij ) is established according to the training data T _{r n×k} in preparation for training the neural network; establish a test fault feature matrix _Re (∑)=(r _ij ) _n×k according to the test data T _e in preparation for testing the neural network expert system after it is established, Where r _ij is the value of the relevant parameter s _i in the actual operation of the diesel engine, the value range of i is {i∈Z|z≤i≤n}, and the value range of j is {j∈Z|1≤j≤k} ; (3) Normalize the fault characteristic matrix R _r (∑) and _Re (∑) in step (2), and obtain the normalized fault characteristic matrix R′ _r (∑)=(r′ _ij ) _n×k and R′ _e (∑)=(r′ _ij ) _n×k , where r′ _ij is the normalized value of the relevant parameter s _i in the actual operation of the diesel engine, and its value range is [0, 1 ], establish training examples and test examples according to the fault feature matrix; (4) Establish a BP neural network and use the training examples in step (3) to train the BP neural network, and use the particle swarm optimization algorithm to accelerate the connection weights ρ _jm ,

and output threshold θ _m ,

In the optimization process, the optimal connection weights ρ _jm ,

and output threshold θ _m ,

Output BP neural network model; (5) Based on the BP neural network obtained in step (4), the connection between it and the expert system is established, that is, the expert knowledge base is established with the output result of the BP neural network, and the structure of the inference engine and the interpreter expert system are established, and the neural network is output. Network expert system model; (6) Use the neural network expert system model obtained in step (5) and the test case in step (3) to test and verify the accuracy of the model. 2. the abnormal diagnosis method of diesel engine exhaust temperature based on neural network expert system according to claim 1, is characterized in that: in the step (1), obtaining relevant operation data comprises obtaining the operation data of 7 groups of relevant measurement points, is respectively diesel engine The data during normal operation and the operating data when there are one of the following faults respectively: the fault types are supercharger damage fault, fuel injection abnormality, intake pipe blockage fault, exhaust valve damage fault, fresh water system abnormality and high-pressure oil pipe Leak failure. 3. the abnormal diagnosis method of diesel engine exhaust temperature based on neural network expert system according to claim 1, is characterized in that: the output result set F in the step (2) comprises: normal operating state f ₁ and fault state, including increasing Compressor damage f ₂ , fuel injection abnormality f ₃ , intake pipe blockage f ₄ , exhaust valve damage f ₅ , fresh water system abnormality f ₆ , high-pressure fuel pipe leakage f ₇ , that is, F={f ₁ , f ₂ , f ₃ , f ₄ , f ₅ , f ₆ , f ₇ }; The input parameter set S includes: exhaust gas temperature s ₁ , turbocharger speed s ₂ , fuel injection status signal s ₃ , air cooler outlet air pressure s ₄ , exhaust valve status signal s ₅ , fresh water inlet pressure s ₆ , high pressure fuel oil Oil leakage level s ₇ , that is, S={s ₁ , s ₂ , s ₃ , s ₄ , s ₅ , s ₆ , s ₇ }; The forms of the fault characteristic matrix R _r (∑) and _Re (∑) are both:

The rows represent seven typical states that may occur in the diesel engine system, and the columns represent seven relevant operating parameters of the diesel engine that can be monitored. 4. the abnormal diagnosis method of diesel engine exhaust temperature based on neural network expert system according to claim 1, is characterized in that: in step (3), fault feature matrix R _r (∑) and R _e (∑) are normalized Normalization processing, that is, normalizing each column of the fault characteristic matrix R _r (∑) and _Re (∑) with the maximum and minimum method:

Among them, i (i=1, ..., 7) is the state type identification, j (j = 1, ..., 7) is the parameter type identification, the value r' _ij after normalization is in [0, 1], the fuel injection state Signal s ₃ and exhaust valve status signal s ₅ are 1 in normal state and 0 in abnormal state; Establish a training example or a test example, that is, take a row of the normalized fault feature matrix as the first dimension of the training example or test example, that is, the input vector, set the state corresponding to the row to 1, and set the other states to 0 , constituting the second dimension of the training example or test example, that is, the expected output vector, a total of 7 training examples are established; taking the intake pipe blockage fault f ₄ as an example, the forms of the training example and the test example are as follows: (x ₄ , y ₄ )=((r′ ₄₁ , r′ ₄₂ , r′ ₄₃ , r′ ₄₄ , r′ ₄₅ , r′ ₄₆ , r′ ₄₇ ) ^T , (0,0,0,1,0 , 0, 0) ^T ). 5. the abnormal diagnosis method of diesel engine exhaust temperature based on neural network expert system according to claim 1, is characterized in that: use particle swarm optimization algorithm to optimize BP neural network in step (4) comprises the following steps: (a) Initialize the BP neural network model: Using a neural network structure with one hidden layer, the number k of neurons in the input layer, the number n of neurons in the output layer, and the number q of neurons in the hidden layer are determined by the following formula:

Among them, t is a constant, t∈[0,10]; Set the input of the mth (1≤m≤q) neuron node of the hidden layer as:

Among them, _ρjm is the connection weight between the jth neuron node in the input layer and the mth neuron node in the hidden layer, and r′ _ij is the normalized value of the jth input parameter in the state i; Set the output of the mth neuron node in the hidden layer as:

in

is the output threshold of the mth neuron node in the hidden layer; Set the i-th neuron node of the output layer, that is, the input of the i-th sub-vector in the output vector is:

in

is the connection weight between the mth neuron node of the hidden layer and the ith neuron node of the output layer, b _h is the output of the mth hidden layer neuron node; Set the i-th neuron node of the output layer, that is, the value of the i-th sub-vector in the output vector is:

Where θ _i is the output threshold of the i-th neuron node in the output layer; (b) Initialize the particle swarm. Set two particle swarm stop iteration conditions, the first is the upper limit of the number of iterations, and the second is that the confidence of the BP neural network reaches 95%: The formula for setting the speed and position update of each particle is:

in

is the component of the velocity of the k+1th generation particle u on the d-th dimension,

is the component of the d-th dimension of the individual extremum of particle u of the kth generation,

is the component of the d-th dimension of the global extremum of the k-th generation particle,

is the component of the position of the kth generation particle u _on the d dimension, c1 and c2 are learning factors, ω is the inertia weight, ζ and η are random numbers on (0, ₁ ); Set the inertia weights to:

k is the number of iterations, k _max is the maximum number of iterations, in the initial state ω=ω _max =1, ω _min =0.4; The fitness function of the particles is set as:

Among them, X is the position of a particle in the particle swarm, n is the number of training examples,

Indicates the output result of the particle under the first training example, and y _i is the expected output result; The particle velocity is initialized, and the value is a uniformly distributed random number between [v _min , v _max ], which is [-0.05, 0.05] in the present invention. Take the connection weight ρ _ij in the BP neural network,

and output threshold θ _m ,

As the position component of each particle in the particle swarm, the position of the uth particle in the particle swarm is:

Among them, x _us is the sth position component of the uth particle, and the value range is [x _min , x _max ]. When a certain position component exceeds the upper or lower bound value, this component is approximately taken as the boundary value The value range of s is {s∈Z|1≤s≤q(k+n+1)+n}, that is, the value range of s is {s∈Z|1≤s≤67}; (c) If the current fitness function value of the particle is better than the individual historical optimal value, update the current position of the particle to the individual historical optimal position; if the individual historical optimal position of the particle is better than the global optimal position, then Update the global optimum of the particle to the current individual historical optimum; (d) Update the position and velocity of each particle according to the particle swarm optimization algorithm; (e) If any one of the stop iteration conditions is met, stop the iteration, get the optimal initial weight and threshold of the BP network, and output the trained BP neural network. 6. the abnormal diagnosis method of diesel engine exhaust temperature based on neural network expert system according to claim 1, is characterized in that: in step (5), based on BP neural network model, the method for setting up its connection with expert system is: Investigate the hidden layer neurons pointing to the output of the BP neural network, and find out the positive weighted input value of the output layer, which has the following characteristics:

At the same time, examine the input layer neurons pointing to this hidden layer neuron, and also find out its positive weighted input value to form a rule: IF expression 1 AND expression 2 AND...AND expression n THEN conclusion The specific form of the expression n in the rule is that the output of a neuron in the input layer is positive, that is, there is a certain symptom, and the specific form of the conclusion in the rule is the fault type; the following rules are obtained for the abnormal temperature of the exhaust gas of the diesel engine: IF exhaust temperature > 550°C AND supercharger speed reduction THEN supercharger damage IF exhaust temperature > 550°C AND air cooler outlet pressure decreases THEN intake pipe is clogged IF exhaust temperature > 550°C AND abnormal fuel injection sign = 1 THEN abnormal fuel injection IF Exhaust temperature > 550°C AND Exhaust valve damaged flag = 1 THEN Exhaust valve damaged IF exhaust temperature > 550°C AND fresh water inlet pressure decreases THEN fresh water system is abnormal IF exhaust temperature is low AND high pressure fuel pipe leaks oil level rises THEN high pressure fuel pipe leaks Store such rules in the knowledge base to form the knowledge base of the neural network expert system, and build an expert system based on this knowledge base. 7. the abnormal diagnosis method of diesel engine exhaust temperature based on neural network expert system according to claim 1, is characterized in that: in the step (6), test neural network expert system flow process is: (i) Receive the alarm signal of abnormal exhaust temperature of the diesel engine; (ii) The inference engine of the expert system refers to the knowledge base, performs fault diagnosis for the abnormal data of the associated measurement points, and calls the neural network module; (iii) If the confidence degree of the diagnosis result of the neural network is greater than 0.95, it will be regarded as the final diagnosis result of the neural network expert system, and the diagnosis result will be explained by the expert system; (iv) If the output confidence of the neural network is not greater than 0.95, then output the diagnosis result of the expert system; (v) If neither the expert system nor the neural network can diagnose the fault, that is, there is a fault that cannot establish the corresponding relationship between symptoms and causes at this stage, then pack the data, start the learning mode of the neural network expert system, and let the expert name the fault , adjust the neural network topology, increase the output nodes, and use the packaged offline data to train the neural network, and update the trained neural network to the neural network expert system.

Images