CN108681248A - A kind of autonomous learning fault diagnosis system that parameter is optimal - Google Patents

Abstract

The invention discloses a self-learning fault diagnosis system with optimal parameters, which is used for fault diagnosis of the Tennessee Eastman process, including a data preprocessing module, a principal component analysis module, a self-learning module and a group intelligence algorithm module. The present invention performs fault diagnosis on important parameter indexes of the Tennessee Eastman chemical process, overcomes the shortcomings of the existing chemical fault diagnosis technology instrument prediction accuracy is not high, and is easily affected by human factors, and utilizes the strong global search ability of the ant colony algorithm and The local search ability quickly finds the optimal parameters of the support vector machine, and a self-learning fault diagnosis system with better diagnostic effect under small sample conditions and easy to obtain optimal parameters is proposed.

Description

A Self-learning Fault Diagnosis System with Optimal Parameters

technical field

The invention relates to the field of fault diagnosis, machine learning and swarm intelligence optimization algorithm, in particular to a Tennessee Eastman process chemical fault diagnosis system combining machine learning and swarm intelligence optimization algorithm.

Background technique

With the rapid development of big data in recent years, big data itself has become an important supporting part of artificial intelligence. Obtaining the information hidden in the data we need has become a new field of scientific research. The use of data for complex industrial process detection has been rapidly developed in the field of industrial fault detection. This technology can effectively reduce the occurrence of industrial production process faults, thereby minimizing the losses caused by faults to enterprises and the country. Chemical enterprises have always been a type of productive enterprise with a high incidence of safety accidents, which have always caused relatively large losses of life and property to local governments and people; for example, on April 3, 2016, a chemical company named Lianhua Technology in Dezhou City, Shandong Province A serious explosion accident occurred in the factory. The explosion caused 7 injuries and caused varying degrees of damage to the surrounding buildings and buildings. Since there are many kinds of chemical raw materials in the chemical plant, the explosion caused a variety of chemical production raw materials to leak, which polluted to a certain extent. local natural environment. On April 20, 2016, an oil processing facility affiliated to Mexico's National Petroleum Corporation exploded. The company involved said that "leakage" was the main factor that triggered the explosion, but did not clearly specify what kind of material leaked. The specific accident The cause needs to be further ascertained. The accident caused a total of 32 deaths and 130 injuries. In February 2017, an explosion occurred in Hengxing Chemical Company in Anhui Province. No casualties were caused in this accident. After the investigation and analysis by the relevant departments, the cause of the accident was initially found to be the misoperation of the steam pipeline valve. The above cases are only part of the accidents that have occurred in the recent period. It can be seen that the probability of safety accidents in the chemical production process is higher than that of other industries; just like this, how to avoid such safety accidents to the greatest extent is particularly important. Carrying out research on fault detection and diagnosis of chemical processes is precisely to reduce the occurrence of such fault accidents, so as to achieve safe production and strive for more economic benefits for the country. Therefore, research on fault detection methods based on data-driven has become an One of the hot issues in the field.

Contents of the invention

In order to overcome the disadvantages of low prediction accuracy and being easily affected by human factors in the existing fault diagnosis technology, the purpose of the present invention is to provide a fault diagnosis system with better prediction effect under the condition of small sample size.

The technical solution adopted by the present invention to solve the technical problem is: a self-learning fault diagnosis system with optimal parameters, which is used for fault diagnosis of the Tennessee Eastman process, including a data preprocessing module, a principal component analysis module, and an autonomous learning module and swarm intelligence algorithm module. in:

Data Preprocessing Module: 52 Variables of the Tennessee Eastman Process It is the input of the data preprocessing module. Since each variable has a different unit, in order to prevent errors between data magnitudes caused by different dimensions, all data are first standardized. The normalization formula is as follows:

Among them, mean represents the arithmetic mean of each variable, std represents the standard deviation of each variable, Indicates the value of the input variable, the subscript i indicates the i-th detection, j indicates the j-th dimension variable, and x _ij indicates the value of the input variable after normalization as the input data. The standardized data is S={x _i1 , x _i2 , . . . x _i52 }.

Principal component analysis module: through principal component analysis, the complexity of the system can be reduced without reducing the accuracy of the system. The standardized data S={x _i1 , x _i2 ,...x _i52 } is subjected to principal component analysis, and 85% of the principal components are retained.

Self-learning module: used to establish a diagnostic system, using support vector machine self-learning module:

Among them, J represents the objective function, w represents the module parameter, b represents paranoia, x represents the input data, y represents the output data, and the subscript i represents the i-th data.

For the solution of the above formula, the Lagrangian multiplier α _i is introduced, and the Lagrangian function L is defined as follows, where the superscript T represents the transposition of the matrix, Represents the kernel function map:

According to the KKT condition, the partial derivative of L(w,α _i ,b) with respect to w and b respectively can be obtained as follows:

This module uses the RBF kernel function K with excellent performance and few required parameters. The function is as follows:

Among them, K is the kernel function, denotes the mean value of the input data, and σ denotes the kernel parameter.

Swarm intelligence algorithm module: use the swarm intelligence method ant colony algorithm to optimize the kernel parameter σ in the self-learning module:

(1) Algorithm initialization, construct the initial solution set S=(s ₁ ,s ₂ ,...,s _n ), determine the size m of the ant colony, set the threshold MaxGen of the iteration times of the ant colony optimization algorithm and initialize the ant colony The number of iterations of group optimization k=0;

(2) Calculate the fitness value F _i (i=1,2,...n) corresponding to the solution set S, the larger the fitness value, the better the solution, and then determine that each solution in the solution set is taken as an ant The probability P _i (i=1,2,...n) of the optimal initial solution, where represents the true value of the solution, Indicates the predicted value of the solution, and n indicates the number of samples:

Initialize the ant number i=0 for executing the optimization algorithm;

(3) Ant i selects a solution in S as the initial solution for optimization, and the selection rule is to do roulette selection according to P;

(4) Ant i performs optimization on the basis of the selected initial solution to find a better solution s _i ;

There are 21 faults in the Tennessee Eastman process. The data of different faults are input into the fault diagnosis system optimized by swarm intelligence for training, and the fault diagnosis model is established.

When the data of unknown faults are input into the fault diagnosis system, the diagnostic result display instrument will display the diagnostic results.

The beneficial effects of the present invention are mainly manifested in: the present invention diagnoses the important parameters of the Tennessee Eastman chemical process, overcomes the shortcomings of the existing chemical fault diagnosis technology, the instrument prediction accuracy is not high, and it is easily affected by human factors. The strong global search ability and local search ability of the group algorithm quickly find the optimal parameters of the support vector machine, and a self-learning fault diagnosis system with better diagnostic effect and easy to obtain the optimal parameters under the condition of small samples is proposed.

Description of drawings

Figure 1 is a schematic diagram of the basic structure of a self-learning fault diagnosis system with optimal parameters;

Fig. 2 is the structural representation of the fault diagnosis system module of swarm intelligence optimization;

Figure 3 is a production flow diagram of the Tennessee Eastman process.

Detailed ways

The present invention will be described in detail below according to the accompanying drawings.

Referring to Fig. 1, a self-learning fault diagnosis system with optimal parameters includes Tennessee Eastman process 1, field smart instrument 2 for measuring easily measurable variables, control station 3 for measuring operating variables, and a database for storing data 4. Fault diagnosis system 5 optimized by swarm intelligence and display device 6 for diagnosis results. The on-site intelligent instrument 2 and the control station 3 are connected to the Tennessee Eastman process 1, the on-site intelligent instrument 2 and the control station 3 are connected to the database 4, and the database 4 is connected to the input terminal of the fault diagnosis system 5 optimized by the group intelligence connected, the output end of the group intelligently optimized fault diagnosis system 5 is connected with the diagnosis result display instrument 6 .

Referring to Figure 3, the variables of the Tennessee Eastman process are shown in Table 1.

Table 1: Tennessee Eastman Process Variables

Numbering process variable Numbering process variable 1 Feed A (flow tube 1) 27 Reactor E feed (streamline 6) 2 Feed D (flow tube 2) 28 Feed to Reactor F (flow tube 6) 3 Feed E (flow tube 3) 29 Feed to Reactor A (flow tube 9) 4 Total feed (flow tube 4) 30 Reactor B feed (flow tube 9) 5 Recirculation flow (flow tube 8) 31 Feed to Reactor C (flow tube 9) 6 Reactor feed rate 32 Reactor D feed (flow tube 9) 7 Reactor pressure 33 Feed to Reactor E (flow tube 9) 8 Reactor level 34 Feed to Reactor F (flow tube 9) 9 Reactor temperature 35 Feed to Reactor G (flow tube 9) 10 Displacement (flow tube 9) 36 Reactor H feed (flow tube 9) 11 Gas-liquid separator temperature 37 Stripper D flow rate (flow tube 11) 12 Gas-liquid separator liquid level 38 Stripper E flow rate (streamline 11) 13 Gas-liquid separator temperature 39 Stripping column F flow rate (flow tube 11) 14 Gas-liquid separator bottom flow (stream 10) 40 Stripper G flow rate (flow tube 11) 15 Stripper level 41 Stripper H flow rate (flow pipe 11) 16 Stripper pressure 42 Dfeed rate 17 Bottom flow of stripper (flow pipe 11) 43 Efeed rate 18 Stripper temperature 44 A feed rate 19 Stripper steam flow 45 total feed rate 20 Compressor power 46 compressor recirculation valve twenty one Reactor cooling water outlet temperature 47 discharge valve twenty two Separator cooling water outlet temperature 48 Separation tank tank liquid flow rate twenty three Reactor A feed (streamline 6) 49 Stripper Liquid Product Flow twenty four Reactor B feed (streamline 6) 50 Stripper water flow 25 Reactor C feed (streamline 6) 51 Reactor cooling water flow 26 Reactor D feed (streamline 6) 52 Condenser cooling water flow

The Tennessee Eastman process data are used as input variables for the swarm intelligence optimized fault diagnosis system 5 . Obtained by manual sampling analysis, collected once every 4 hours for analysis.

With reference to Fig. 2, the fault diagnosis system 5 of described swarm intelligence optimization also includes:

Data Preprocessing Module 7: 52 Variables of the Tennessee Eastman Process It is the input of the data preprocessing module. Since each variable has a different unit, in order to prevent errors between data magnitudes caused by different dimensions, all data are first standardized. The normalization formula is as follows:

Among them, mean represents the arithmetic mean of each variable, std represents the standard deviation of each variable, Indicates the value of the input variable, the subscript i indicates the i-th detection, j indicates the j-th dimension variable, and x _ij indicates the value of the input variable after normalization as the input data. The standardized data is S={x _i1 , x _i2 , . . . x _i52 }.

Principal component analysis module 8: use principal component analysis to ensure that the complexity of the system is reduced without reducing the accuracy of the system. The standardized data S={x _i1 , x _i2 ,...x _i52 } is subjected to principal component analysis, and 85% of the principal components are retained.

Self-learning module 9: used to establish a diagnostic system, using support vector machine self-learning module:

Among them, J represents the objective function, w represents the module parameter, b represents paranoia, x represents the input data, y represents the output data, and the subscript i represents the i-th data.

For the solution of the above formula, the Lagrangian multiplier α _i is introduced, and the Lagrangian function L is defined as follows, where the superscript T represents the transposition of the matrix, Represents the kernel function map:

According to the KKT condition, the partial derivative of L(w,α _i ,b) with respect to w and b respectively can be obtained as follows:

This module uses the RBF kernel function K with excellent performance and few required parameters. The function is as follows:

Among them, K is the kernel function, denotes the mean value of the input data, and σ denotes the kernel parameter.

Swarm Intelligence Algorithm Module 10: Use the swarm intelligence method ant colony algorithm to optimize the kernel parameter σ in the autonomous learning module:

(1) Algorithm initialization, construct the initial solution set S=(s ₁ ,s ₂ ,...,s _n ), determine the size m of the ant colony, set the threshold MaxGen of the iteration times of the ant colony optimization algorithm and initialize the ant colony The number of iterations of group optimization k=0;

(2) Calculate the fitness value F _i (i=1,2,...n) corresponding to the solution set S, the larger the fitness value, the better the solution, and then determine that each solution in the solution set is taken as an ant The probability P _i (i=1,2,...n) of the optimal initial solution, where represents the true value of the solution, Indicates the predicted value of the solution, and n indicates the number of samples:

Initialize the ant number i=0 for executing the optimization algorithm;

(3) Ant i selects a solution in S as the initial solution for optimization, and the selection rule is to do roulette selection according to P;

(4) Ant i performs optimization on the basis of the selected initial solution to find a better solution s _i ;

There are 21 faults in the Tennessee Eastman process, and the data of different faults are input into the swarm intelligence optimized fault diagnosis system 5 for training to establish a fault diagnosis model.

When the data of an unknown fault is input into the fault diagnosis system, the diagnostic result display instrument 6 displays the diagnostic result.

The embodiments of the present invention are used to explain the present invention, rather than to limit the present invention. Within the spirit of the present invention and the protection scope of the claims, any modification and change made to the present invention will fall into the protection scope of the present invention.

Claims (5)

1. a kind of autonomous learning fault diagnosis system that parameter is optimal, for carrying out fault diagnosis to Tennessee Yi Siman processes, It is characterized in that, it includes data preprocessing module, principal component analysis module, autonomous learning module and swarm intelligence algorithm mould Block.

2. the optimal autonomous learning fault diagnosis system of parameter according to claim 1, which is characterized in that the data are located in advance The input for managing module is 52 variables of Tennessee Yi Siman processesSince each variable is different Unit, in order to prevent different dimensions cause the error between data magnitude, first all data are standardized, mark Standardization formula is as follows：

Wherein, mean indicates that the arithmetic mean of instantaneous value of each variable, std indicate the standard deviation of each variable,Indicate the value of input variable, Subscript i indicates that ith detection, j indicate that jth ties up variable, x respectively_ijIndicate the value of input variable after standardizing as input data. Data after standardization are S={ x_i1,x_i2,...x_i52}。

3. the optimal autonomous learning fault diagnosis system of parameter according to claim 1, which is characterized in that the principal component point Analysis module ensures the complexity of the reduction system in the case where not reducing system accuracy by principal component analysis.After standardizing Data S={ x_i1,x_i2,...x_i52Principal component analysis is carried out, retain 85% main component.

4. the optimal autonomous learning fault diagnosis system of parameter according to claim 1, which is characterized in that the autonomous learning Module is for establishing diagnostic system, using support vector machines autonomous learning module：

Wherein J indicates that object function, w representation modules parameter, b indicate that bigoted, x indicates that input data, y indicate output data, under It marks i and indicates i-th of data.

Solution for above-mentioned formula introduces Lagrange multiplier α_i, definition LagrangianL is as follows, wherein subscript T tables Show the transposition of matrix,Indicate kernel function mapping：

According to KKT conditions, by L (w, α_i, b) and local derviation is asked to w, b respectively, following formula can be obtained：

For the module using the few RBF kernel function K of excellent performance, required parameter, function is as follows：

Wherein, K is kernel function,Indicate that the average value of input data, σ indicate nuclear parameter.

5. the optimal autonomous learning fault diagnosis system of parameter according to claim 1, which is characterized in that the colony intelligence is calculated Method module optimizes autonomous learning mould nuclear parameter σ in the block using swarm intelligence ant group algorithm, specially：

(1) algorithm initialization constructs initial disaggregation S=(s₁,s₂,...,s_n), determine that the size m of ant colony, setting ant colony are sought The threshold value MaxGen of excellent algorithm iteration number and the iterations serial number k=0 for initializing ant optimization；

(2) the corresponding fitness value F of disaggregation S are calculated_i(i=1,2 ... n), fitness value is bigger to represent Xie Yuehao, then determines The each solution of solution concentration is fetched into the probability P of the initial solution as ant optimizing_i(i=1,2 ... n), whereinIndicate the true of solution Value,Indicate that the predicted value of solution, n indicate number of samples：

Initialization executes the ant number i=0 of optimizing algorithm；

(3) ant i chooses initial solution of the solution as optimizing in S, and selection rule is to do wheel disc choosing according to P；

(4) ant i carries out optimizing on the basis of the initial solution of selection, finds preferably solution s_i；

Tennessee Yi Siman processes share 21 failures, and the data of different faults are input to the fault diagnosis system of colony intelligence optimization It is trained in system, establishes fault diagnosis model.

When the data of unknown failure are input to this fault diagnosis system, diagnostic result display instrument shows diagnostic result.