CN111767657A - Nuclear power system fault diagnosis method and system - Google Patents
Nuclear power system fault diagnosis method and system Download PDFInfo
- Publication number
- CN111767657A CN111767657A CN202010654755.9A CN202010654755A CN111767657A CN 111767657 A CN111767657 A CN 111767657A CN 202010654755 A CN202010654755 A CN 202010654755A CN 111767657 A CN111767657 A CN 111767657A
- Authority
- CN
- China
- Prior art keywords
- fault
- support vector
- vector machine
- machine model
- nuclear power
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to a nuclear power system fault diagnosis method and system. According to the nuclear power system fault diagnosis method and system, the fault category of each subsystem in the nuclear power system and the fault occurrence probability corresponding to the fault category can be accurately obtained by adopting the trained nonlinear support vector machine model. And after a normalization coefficient is determined according to the fault occurrence probability in the constructed two-dimensional fault probability matrix, the fault probability value corresponding to each fault is determined according to the normalization coefficient. And then, comparing the difference value between the first large fault probability value and the second large fault probability value in the probability values after descending order with a set threshold value to obtain a fault diagnosis result of the nuclear power system, so that the whole nuclear power system fault diagnosis method and system improve the diagnosis accuracy and the adaptability.
Description
Technical Field
The invention relates to the field of nuclear power system detection, in particular to a nuclear power system fault diagnosis method and system.
Background
Unlike other power plants, nuclear power systems are complex in structure, have radioactive hazards, and have extremely high requirements for safety. Meanwhile, the space of the nuclear power system for the ship is limited, the equipment arrangement is relatively compact, and excessive maintenance personnel cannot be carried, so that the requirement on the reliability of the nuclear power system and key equipment is very high. In addition, the operating environment of the marine nuclear power system is severe, and key equipment of the system continuously works for a long time, so that faults are very easy to occur, if the equipment fails and cannot be found and maintained in time, serious radioactive consequences can be caused, and the life safety of a crew is critical.
At present, the traditional threshold analysis and manual experience are mostly adopted for judging the complex fault analysis and diagnosis technology of the nuclear power system and key equipment. However, these conventional techniques do not fully accommodate the reliability requirements of complex systems and critical equipment. With the continuous development of artificial intelligence technology and big data theory, the accumulation of a large amount of operation data of a marine nuclear power system and application experience in other fields, the operation state of the system is monitored by adopting a plurality of high-efficiency and accurate artificial intelligence technologies, fault diagnosis is rapidly and accurately carried out, a set of characteristic identification and diagnosis evaluation method system under complex faults is formed, the maintenance guarantee capability of the nuclear power system and key equipment can be effectively improved, the maintenance cost is reduced, and autonomous guarantee is realized.
In 1967, the mechanical failure prevention group was established by the naval research institute in the united states, from which the research work of failure diagnosis technology began, and then the research and application of failure diagnosis technology gradually spread all over the world. The establishment of the british association of machine health and condition monitoring in the end of the 60's 20 th century further pushed the development of fault diagnosis technology. Subsequently, various countries in europe have also developed related research on condition monitoring and fault diagnosis technologies, and developed diagnostic technology systems with respective features. The fault diagnosis technology of Japan starts to start in the middle of the 70 s, and by learning and using for reference of research of various countries in the world, continuous improvement and improvement, the fault diagnosis technology of Japan in civil industries such as steel production, railway operation, chemical process and the like is mature at present; the research related to the Chinese fault diagnosis technology starts in the early 80 s, and a relatively perfect theoretical system is formed at present.
As shown in fig. 1, the fault diagnosis method can be classified into a quantitative model-based method, a qualitative model-based method, and a historical data-based method. In the nuclear field, research and development of 200MW nuclear heating plant fault diagnosis system are carried out by Qinghua university. The Harbin engineering university designs and develops a nuclear power system operation support system, which comprises the functions of state monitoring, alarm analysis, fault diagnosis, emergency operation guidance and the like. An accident diagnosis counseling system for fault diagnosis of a nuclear power plant was developed by the korean science and technology institute.
In addition, the application of the neural network in the identification of the nuclear power plant starting event is researched by T.V.Santosh, and several learning algorithms are contrastively analyzed, so that the conclusion that the elastic BP algorithm is the optimal learning algorithm is obtained. Marzio Marseguerra studies fuzzy recognition of transients in nuclear power plants, and E.Zio proposes an improved fuzzy clustering method for classifying and recognizing transients in nuclear power plant equipment. Carla Regina Gome et al used a Gaussian radial basis function neural network to analyze pressurized water reactor power plant accidents. The Sinuhe Martinez-Martinez application detects the reactor core assembly blockage fault of the sodium-cooled fast reactor based on the artificial neural network strategy.
C. A multi-layer neural network of the 'jump' type is provided, and two neural networks are used for dynamically identifying and verifying the identification result respectively. KunMo proposes a framework of a two-layer dynamic neural network, the first layer identifying the type of fault and the second layer determining the location and extent of the fault. Sercat marker proposed the use of Elman neural networks to monitor anomalies in high temperature gas cooled reactors. Jose studies the fault identification of particle swarm optimization algorithms in nuclear power plants.
The Xindongdong of Harbin engineering university researches the application of a BP Neural Network in nuclear power system fault diagnosis, improves a learning algorithm to ensure the rapidness and accuracy of diagnosis, introduces a Radial Basis Function (RBF) Neural Network and a Fuzzy Neural Network (FNN) into a nuclear power system fault diagnosis system for Liu Yongkun, and uses the FNN for local diagnosis and data fusion for global diagnosis, and provides a distributed diagnosis strategy. And carrying out system-level fault diagnosis by adopting a genetic algorithm in Dunwei. An Elman neural network is researched in real-time prediction method research on characteristic parameter corresponding values in high-pressure water supply system fault diagnosis by Malaya jade, Queenxia and the like of North China Power university, an ant colony algorithm is researched in application of fault diagnosis by Whitemadam, and a nuclear power system fault diagnosis algorithm based on an invasive weed algorithm is researched by Changqiang.
However, the above prior art provides a method which cannot accurately determine a complex accident in a nuclear power system.
Disclosure of Invention
The invention aims to provide a nuclear power system fault diagnosis method and system to improve the accuracy of determining complex accidents in a nuclear power system.
In order to achieve the purpose, the invention provides the following scheme:
a nuclear power system fault diagnostic method comprising:
acquiring a plurality of trained nonlinear support vector machine models which are the same as the number of subsystems to be diagnosed in the nuclear power system;
acquiring operation data of each subsystem to be diagnosed by using a sensor;
outputting the fault category of each subsystem to be diagnosed and the fault occurrence probability corresponding to the fault category by adopting the nonlinear support vector machine model according to the operation data;
taking the number of the nonlinear support vector machine models as a row and the occurrence probability of the faults as a column, and forming a two-dimensional fault probability matrix by the result output by each nonlinear support vector machine model;
determining a normalization coefficient for the fault occurrence probability in the two-dimensional fault probability matrix;
determining a fault probability value corresponding to each fault according to the normalization coefficient, and performing descending order arrangement on all fault probability values;
when the difference between the first large fault probability value and the second large fault probability value after descending order is smaller than a set threshold, the probability of occurrence of the two faults is determined to be large, and the two faults are correspondingly diagnosed as 'unrecognizable'; and if the difference between the first large fault probability value and the second large fault probability value is larger than a set threshold value, determining the fault type corresponding to the first large fault probability value as the current fault state.
Preferably, the training process of the non-linear support vector machine model includes:
acquiring historical operating data of the same reactor type nuclear power system;
simulating the historical operation data by adopting a full-range simulator to obtain simulation data of various single accident data and complex accidents;
carrying out calibration sampling on the simulation data to obtain the fault category in the same reactor type nuclear power system and the fault occurrence probability corresponding to the fault category;
and taking the historical operating data, the fault category and the fault occurrence probability corresponding to the fault category as a training data sample pair, and training the nonlinear support vector machine model to obtain a trained nonlinear support vector machine model.
Preferably, the training of the nonlinear support vector machine model with the historical operating data, the fault category, and the fault occurrence probability corresponding to the fault category as a training data sample pair to obtain the trained nonlinear support vector machine model specifically includes:
inputting the training sample pair into the nonlinear support vector machine model to obtain the classification accuracy of the nonlinear support vector machine model;
taking the classification accuracy as a fitness function, and performing self-adaptive optimization on the parameters to be optimized by adopting a particle swarm algorithm; the parameters to be optimized are penalty factors and the width of a kernel function in the nonlinear support vector machine model;
and taking the optimized punishment factor and the width of the kernel function as a new punishment factor and the width of a new kernel function in the nonlinear support vector machine model to obtain the trained nonlinear support vector machine model.
Preferably, the classification accuracy is used as a fitness function, and a particle swarm algorithm is adopted to perform adaptive optimization on the parameters to be optimized, and the method specifically comprises the following steps:
initializing the inertial weight, cognitive learning factors and social learning factors of the particle swarm algorithm, and determining the fitness of the initialized initial population;
in the iterative process of the particle swarm optimization, random variation is carried out on the particle swarm by adopting a self-adaptive large-scale variation guarantee algorithm so as to determine global optimal particles and the fitness of the global optimal particles in each iterative process;
in the continuous iteration process, when the accumulated times of the fitness of the global optimal particles is greater than or equal to the set fitness threshold value and greater than or equal to the set value, the self-adaptive optimization of the parameters to be optimized is completed.
Preferably, after the acquiring the operation data of each subsystem to be diagnosed by using the sensor, the method further includes:
and normalizing the operation data by using a dispersion standardization method.
A nuclear power system fault diagnostic system comprising:
the nonlinear support vector machine model acquisition module is used for acquiring a plurality of trained nonlinear support vector machine models which are the same as the number of subsystems to be diagnosed in the nuclear power system;
the operation data acquisition module is used for acquiring the operation data of each subsystem to be diagnosed by adopting a sensor;
the fault type and fault occurrence probability determining module is used for outputting the fault type in each subsystem to be diagnosed and the fault occurrence probability corresponding to the fault type by adopting the nonlinear support vector machine model according to the operation data;
the two-dimensional fault probability matrix construction module is used for forming a two-dimensional fault probability matrix by using the number of the nonlinear support vector machine models as rows and the fault occurrence probability as columns and outputting results of each nonlinear support vector machine model;
the normalization coefficient determining module is used for determining a normalization coefficient for the fault occurrence probability in the two-dimensional fault probability matrix;
the descending order ranking module is used for determining the fault probability value corresponding to each fault according to the normalization coefficient and carrying out descending order ranking on all the fault probability values;
a fault determining module, configured to determine that the two kinds of faults have high probability and are correspondingly diagnosed as "unrecognizable" when a difference between the first large fault probability value and the second large fault probability value after the descending order is smaller than a set threshold; and if the difference between the first large fault probability value and the second large fault probability value is larger than a set threshold value, determining the fault type corresponding to the first large fault probability value as the current fault state.
Preferably, the method further comprises the following steps:
the historical operating data acquisition module is used for acquiring historical operating data of the nuclear power system of the same reactor type;
the simulation data determining module is used for simulating the historical operation data by adopting a full-range simulator to obtain simulation data of various single accident data and complex accidents;
the calibration sampling module is used for performing calibration sampling on the simulation data to obtain the fault category in the same type of reactor type nuclear power system and the fault occurrence probability corresponding to the fault category;
and the nonlinear support vector machine model training module is used for taking the historical operating data, the fault category and the fault occurrence probability corresponding to the fault category as a training data sample pair, and training the nonlinear support vector machine model to obtain a trained nonlinear support vector machine model.
Preferably, the nonlinear support vector machine model training module specifically includes:
a classification accuracy determining unit, configured to input the training sample pair into the nonlinear support vector machine model to obtain a classification accuracy of the nonlinear support vector machine model;
the parameter optimizing unit is used for taking the classification accuracy as a fitness function and carrying out self-adaptive optimization on the parameters to be optimized by adopting a particle swarm algorithm; the parameters to be optimized are penalty factors and the width of a kernel function in the nonlinear support vector machine model;
and the nonlinear support vector machine model training unit is used for taking the optimized punishment factor and the width of the kernel function as a new punishment factor and a new width of the kernel function in the nonlinear support vector machine model to obtain the trained nonlinear support vector machine model.
Preferably, the parameter optimizing unit specifically includes:
the initial population fitness determining subunit is used for initializing the inertial weight, the cognitive learning factor and the social learning factor of the particle swarm algorithm and determining the fitness of the initialized initial population;
the global optimal particle fitness determining subunit is used for performing random variation on the particle swarm by adopting a self-adaptive large-scale variation guarantee algorithm in the iterative process of the particle swarm algorithm so as to determine global optimal particles and the fitness of the global optimal particles in each iterative process;
and the parameter optimizing subunit is used for finishing the self-adaptive optimization of the parameter to be optimized when the accumulated times that the fitness of the global optimal particle is greater than or equal to the set fitness threshold value is greater than or equal to the set value in the continuous iteration process.
Preferably, the system further comprises:
and the normalization processing module is used for performing normalization processing on the operation data by using a dispersion normalization method.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
according to the nuclear power system fault diagnosis method and system provided by the invention, the fault category of each subsystem in the nuclear power system and the fault occurrence probability corresponding to the fault category can be accurately obtained by adopting the trained nonlinear support vector machine model. And after a normalization coefficient is determined according to the fault occurrence probability in the constructed two-dimensional fault probability matrix, the fault probability value corresponding to each fault is determined according to the normalization coefficient. And then, comparing the difference value between the first large fault probability value and the second large fault probability value in the probability values after descending order with a set threshold value to obtain a fault diagnosis result of the nuclear power system, so that the whole nuclear power system fault diagnosis method and system improve the diagnosis accuracy and the adaptability.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 is a diagram of a fault diagnosis classification in the prior art.
FIG. 2 is a flow chart of a nuclear power system fault diagnosis method provided by the present invention;
FIG. 3 is a flow chart of a technique for diagnosing a complex event of a nuclear power system in accordance with an embodiment of the present invention;
FIG. 4 is a flowchart illustrating adaptive optimization of parameters to be optimized by using a particle swarm optimization in the embodiment of the present invention;
fig. 5 is a schematic structural diagram of a nuclear power system fault diagnosis system provided by the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a nuclear power system fault diagnosis method and system to improve the accuracy of determining complex accidents in a nuclear power system.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 2 is a flowchart of a nuclear power system fault diagnosis method provided by the present invention. As shown in fig. 2, the method for diagnosing a fault of a nuclear power system according to the present invention includes:
step 100: and acquiring a plurality of trained nonlinear support vector machine models which are the same as the number of subsystems to be diagnosed in the nuclear power system.
Step 101: and acquiring the operation data of each subsystem to be diagnosed by adopting a sensor.
Step 102: and outputting the fault category of each subsystem to be diagnosed and the fault occurrence probability corresponding to the fault category by adopting the nonlinear support vector machine model according to the operation data.
Step 103: and taking the number of the nonlinear support vector machine models as a row and the fault occurrence probability as a column, and forming a two-dimensional fault probability matrix by using the result output by each nonlinear support vector machine model.
Step 104: and determining a normalization coefficient for the fault occurrence probability in the two-dimensional fault probability matrix.
Step 105: and determining the fault probability value corresponding to each fault according to the normalization coefficient, and performing descending order arrangement on all the fault probability values.
Step 106: when the difference between the first large fault probability value and the second large fault probability value after descending order is smaller than a set threshold, the probability of occurrence of the two faults is determined to be large, and the two faults are correspondingly diagnosed as 'unrecognizable'; and if the difference between the first large fault probability value and the second large fault probability value is larger than a set threshold value, determining the fault type corresponding to the first large fault probability value as the current fault state.
In order to further improve the accuracy of the output result of the non-linear support vector machine model, in the present invention, before step 100, the non-linear support vector machine model needs to be trained, and the training process includes:
acquiring historical operating data of the same reactor type nuclear power system; and performing classified management on the collected operation data in the computer according to the subsystem to which the sensor belongs.
And simulating the historical operation data by adopting a full-range simulator to obtain simulation data of various single accidents and complex accidents.
Carrying out calibration sampling on the simulation data to obtain the fault category in the same reactor type nuclear power system and the fault occurrence probability corresponding to the fault category; the calibration sampling of the simulation data specifically comprises the following steps: and marking the simulation data so as to set the normal state and different fault states as different labels.
And taking the historical operating data, the fault category and the fault occurrence probability corresponding to the fault category as a training data sample pair, and training the nonlinear support vector machine model to obtain a trained nonlinear support vector machine model. And, the step may further include:
inputting the training sample pair into the nonlinear support vector machine model to obtain the classification accuracy of the nonlinear support vector machine model.
Taking the classification accuracy as a fitness function, and performing self-adaptive optimization on the parameters to be optimized by adopting a particle swarm algorithm; the parameters to be optimized are penalty factors and the width of a kernel function in the nonlinear support vector machine model.
The above-mentioned classification accuracy is regarded as the fitness function, adopts the particle swarm algorithm to treat optimizing parameter and carries out the self-adaptation and optimizes, specifically includes:
initializing the inertial weight, the cognitive learning factor and the social learning factor of the particle swarm algorithm, and determining the fitness of the initialized initial population.
In the iterative process of the particle swarm optimization, random variation is carried out on the particle swarm by adopting a self-adaptive large-scale variation guarantee algorithm so as to determine the global optimal particles and the fitness of the global optimal particles in each iterative process.
In the continuous iteration process, when the accumulated times of the fitness of the global optimal particles is greater than or equal to the set fitness threshold value and greater than or equal to the set value, the self-adaptive optimization of the parameters to be optimized is completed. The fitness threshold set in the present invention is preferably 90%, and the set value is preferably 5.
And taking the optimized punishment factor and the width of the kernel function as a new punishment factor and the width of a new kernel function in the nonlinear support vector machine model to obtain the trained nonlinear support vector machine model.
In order to avoid the influence of the data with inconsistent dimension, too large and too small on the output result, after the historical operating data is obtained in step 101, the operating data can be normalized by using a dispersion normalization method. The process is embodied in that all data values of the same parameter are mapped to [0,1]]In the meantime. The mapping function is: x is the number of*(x-min)/(max-min), where max is the sample maximum and min is the sample minimum.
The following provides a specific embodiment to further illustrate the solution of the present invention, and in the specific embodiment of the present invention, the non-linear support vector machine model with the radial basis function as the kernel function is taken as an example for explanation, and in a specific application, the solution of the present invention is also applicable to the non-linear support vector machine models of other kernel functions.
FIG. 3 is a flow chart of a technique for diagnosing a complex event of a nuclear power system in accordance with an embodiment of the present invention; as shown in fig. 3, the process of the method for diagnosing a fault of a nuclear power system based on a nonlinear support vector machine model with a kernel function as a radial basis function specifically includes:
step 1: the method comprises the steps of collecting and storing actual operation data of the nuclear power system of the same reactor type and operation data of a full-range simulator under various single accidents and complex accidents. Wherein the operational data includes: the pressure of a pressure stabilizer in a reactor coolant system, the temperature of a fluctuation pipe, the flow rate of an outlet on the primary side of a steam generator, the temperature of an inlet and an outlet of a reactor core, the water level of the secondary side of the steam generator, the feed water temperature and the feed water flow rate, the steam yield and the steam temperature, the upper charge flow rate and the lower discharge flow rate of a chemical volume system, the water level of a volume control box and the like.
Step 2: collected operation data are managed in a computer in a classified manner according to subsystems to which the sensors belong, actual historical operation data and simulation data in each subsystem are labeled, and different labels can be set for normal states and different fault states. The reactor coolant system in the nuclear power system is used as a subsystem, the first loop auxiliary system is used as a subsystem, and the parameter of the second loop system is used as a subsystem.
And step 3: and (3) normalizing all the data in the step (2) according to the same standard, so as to avoid the influence of inconsistent dimension pairs and overlarge and undersize data on data training. All data values of the same parameter are mapped between 0,1 using a dispersion normalization method.
And 4, step 4: and establishing a corresponding number of nonlinear support vector machine models based on penalty factors according to the number of systems to be diagnosed and analyzed in the nuclear power system.
On each classification algorithm in the nonlinear support vector machine models with the radial basis function as the kernel function, each nonlinear support vector machine model selects a "one-to-many" algorithm, namely for N classes of classification problems: firstly, establishing N two-classification sub-classifiers, wherein the ith sub-classifier takes the ith class as a positive class, all the other classes are negative classes, and finally, each nonlinear support vector machine model respectively outputs the occurrence probability of the corresponding fault type and carries out the sorting from large to small according to the probability.
And 5: grouping the normalized data in the step 3 according to the number of the subsystems of the nuclear power system, wherein each nonlinear support vector machine model in the step 4 corresponds to the measurement point acquired in the corresponding subsystem respectively; and respectively transmitting the normal data of different measuring points, the abnormal data of different faults and different fault degrees and class labels thereof to corresponding nonlinear support vector machine models so as to train the nonlinear support vector machine models.
The objective of the training of the nonlinear support vector machine model is to find the optimum value of the width of the kernel function (set as parameter g) and the penalty factor (set as parameter c) for adjusting the classification bandwidth of the support vector machine in each support vector machine. If the experience setting is adopted, the uncertainty is too large, and the accurate classification of the support vector machine is difficult to ensure; if a grid search algorithm and the like are adopted, the search accuracy is difficult to guarantee if the search effect is too low, and the best effect cannot be obtained. Therefore, the invention adopts the improved particle swarm optimization to carry out the self-adaptive optimization of the parameters c and g in the support vector machine, and the flow of the improved particle swarm optimization is shown in figure 4.
Step 6: for the nonlinear support vector machine model corresponding to the first subsystem, the training data and the class labels of the training data corresponding to the measuring points are respectively transmitted to the nonlinear support vector machine model, the output classification accuracy is used as a fitness function of the improved particle swarm algorithm, and the parameters c and g in the support vector machine are respectively used as two parameters to be optimized for improving the particle swarm.
And 7: initializing parameters such as inertia weight, learning factors and the like of the particle swarm, and calculating the fitness of the initial population.
And 8: judging whether the current iteration time reaches the maximum time, if so, transmitting the currently obtained global optimum value back to the nonlinear support vector machine model; and if the iteration time is less than the maximum iteration time, continuously executing particle swarm parameter optimization calculation.
And step 9: and respectively carrying out gradual adjustment on the inertia weight, the cognitive learning factor and the social learning factor in the particle swarm optimization in an iterative process by adopting a nonlinear adjustment algorithm. The step can avoid the mismatching of the linear descending weight and the actual searching process in the basic particle swarm optimization.
The inertia weight, the cognitive learning factor and the social learning factor are respectively adjusted by adopting the following calculation formulas:
x＝a×exp(bk2)×rand(0,1) (1)
a＝wmaxexp(-b) (3)
w is the maximum and minimum of the inertial weight or cognitive or social learning factor, each of which is calculated independently, except that the formula is the same. In the present invention, the maximum value of the inertial weight is preferably 0.9, and the minimum value thereof is preferably 0.4. The maximum and minimum values of the two learning factors are preferably 2 and 0.5, respectively. rand (0,1) is the random number generated between 0 and1, K is the total number of iterations, and K is the current number of iterations. a, b are two coefficients, exp is an exponential function with a natural constant e as the base.
Step 10: and updating the migration speed of the particle swarm according to a speed formula, and updating the speed of the particles according to a position formula to obtain the values of the parameter c and the parameter g in the particle swarm at the new moment. Respectively calculating the fitness and updating the individual extreme value and the global extreme value;
step 11: and (3) carrying out random variation on the particle swarm by adopting a self-adaptive large-scale variation guarantee algorithm, wherein a variation formula is shown as a formula (4), and rand1 respectively represent two independent random numbers between 0 and 1. The particles can be changed to a larger extent, and the particles are well prevented from falling into local optimum.
And then recalculating all the particle fitness values after the variation, obtaining a global optimal particle after comparison, and calculating the fitness value of the particle.
Step 12: if the fitness value is greater than or equal to 90%, counting for 1 time, and if the continuous counting is greater than or equal to 5, transmitting the obtained globally optimal particles to a nonlinear support vector machine model to complete optimization of the parameters c and g; and if the fitness value is less than 90%, returning to the step 8, and repeating the steps 8 to 11 until the termination condition is reached.
Step 13: and (3) repeating the steps 6-12 for each support vector machine model corresponding to other subsystems until each nonlinear support vector machine model meets the termination condition in the step 12, finally obtaining the optimal parameter c and the optimal parameter g corresponding to each nonlinear support vector machine model aiming at the training data, and finishing the training process of the whole nonlinear support vector machine model.
Step 14: in the actual complex fault diagnosis process, preprocessing the abnormal data according to the mode of the step 1-3, and ensuring that the data processing mode is completely consistent with the training data; and then grouping according to the measuring point grouping mode in the step 5, and respectively transmitting the measuring point data to the corresponding nonlinear support vector machine models.
Step 15: and each support vector machine model diagnoses and outputs respective fault diagnosis probability results according to actual measured point data respectively, and displays the results from large to small.
Step 16: because each nonlinear support vector machine model only concerns the abnormality reflected in the corresponding subsystem, in order to comprehensively analyze complex accidents occurring in the nuclear power system and obtain a result with higher accuracy, a data fusion model is established by adopting a D-S evidence theory.
And step 17: and forming a two-dimensional fault probability matrix with the number of rows being the number of the nonlinear support vector machine models and the number of columns being the number of the faults identified during training by using the fault probability results output by the diagnosis of each nonlinear support vector machine model.
Step 18: and substituting the probability value in the two-dimensional fault probability matrix into the following formula (5) to obtain a normalization coefficient cr according to the difficulty degree of actual calculation.
In the formula, AiDiagnosing faults for all possible propositions; mi() The probability distribution function is a support vector machine i in the invention; mi(Ai) Meaning here proposition AiProbability values in the support vector machine i decision results.
Step 19: and (4) substituting the probability value in the fault probability matrix and the normalization coefficient obtained in the step (18) into a formula by using the formula (6) to obtain the probability value of the fault A after fusion.
M (A) is the probability value of the fault A after fusion.
Step 20: and repeating the process of the step 19, sequentially solving the fusion probability values of the other faults, and forming a one-dimensional fault probability matrix by the obtained final probability values.
Step 21: the fused probabilities are the confidence degrees of the fault occurrence, so that the probability of which fault occurs is the largest by sequencing all the fault probabilities from large to small, and when the difference between the two largest fault probabilities is smaller than a threshold value, the two fault probabilities are determined to be large and diagnosed as 'unidentifiable'; if the difference is larger than the threshold value, the fault with the maximum probability is determined as the current fault state.
In the technical scheme provided by the invention, the nonlinear vector machine model can be replaced by machine learning modes such as a decision tree and a back propagation neural network, so that fault identification and fault diagnosis can be carried out. However, since the fault data itself is very limited, the combined data of different faults after a complex accident occurs is more poor, and the machine learning methods such as the decision tree and the back propagation neural network cannot provide a more accurate diagnosis result under the condition of a small quantity of fault samples.
In summary, the diagnostic method provided by the present invention has the following advantages compared to the prior art:
1. the nonlinear support vector machine is used as a base classifier to perform fault diagnosis and mode recognition on each subsystem of the nuclear power system, and the theoretical basis of the nonlinear support vector machine ensures that the fault diagnosis of the method under a small sample can reach the optimal state, so that the nonlinear support vector machine is used as the base classifier to obtain the relatively optimal diagnosis result under the background that the fault data of the nuclear power system is limited.
2. By establishing a nonlinear support vector machine model for each subsystem, the characteristic change condition of each subsystem after the fault occurs can be memorized, and the diagnosis effect and accuracy of the complex accident are improved.
3. Parameter optimization is carried out on the penalty factor of each nonlinear support vector machine model and the width of the kernel function by adopting an improved particle swarm algorithm, the problems that the grid search algorithm is too low in search accuracy or too long in time consumption can be solved, and the diagnosis accuracy is improved.
4. The inertial weight, the cognitive learning factor and the social learning factor of the particle swarm optimization are iteratively adjusted by adopting a nonlinear adjustment algorithm, so that mismatching of the linear descending weight in the basic particle swarm optimization and the actual searching process can be avoided, and the probability of searching the optimal parameters by the particle swarm optimization is further improved.
5. The self-adaptive large-scale variation formula is adopted to ensure that the particles can randomly deviate from the positions of the particles with higher probability in each iteration process, so that the particle swarm can be prevented from being trapped into partial optimization in the process of searching the optimal parameters of the base classifier, and the influence of the particle swarm on the accuracy of diagnosis and analysis is further avoided.
6. Setting optimization termination conditions can ensure that the user can quit in time after finding the optimal solution, reduce training time and avoid unnecessary time resource consumption in the training process.
7. Each nonlinear support vector machine model outputs a probability result instead of directly outputting a classification label, so that the diagnosis results can be sorted according to the probability to assist the subsequent D-S evidence theory in fusing final results.
8. The probability results output by each support vector machine can be fused by adopting a D-S evidence theory, so that more reasonable results with higher diagnosis accuracy can be obtained.
In addition, aiming at the nuclear power system fault diagnosis method, the invention also correspondingly provides a nuclear power system fault diagnosis system. As shown in fig. 5, the system includes: the method comprises a nonlinear support vector machine model acquisition module 200, an operation data acquisition module 201, a fault category and fault occurrence probability determination module 202, a two-dimensional fault probability matrix construction module 203, a normalization coefficient determination module 204, a descending order arrangement module 205 and a fault determination module 206.
The nonlinear support vector machine model obtaining module 200 is configured to obtain a plurality of trained nonlinear support vector machine models, which are the same as the number of subsystems to be diagnosed in the nuclear power system.
The operation data acquiring module 201 is configured to acquire operation data of each subsystem to be diagnosed by using a sensor.
The failure type and failure occurrence probability determining module 202 is configured to output, according to the operation data, a failure type in each subsystem to be diagnosed and a failure occurrence probability corresponding to the failure type using the nonlinear support vector machine model.
The two-dimensional fault probability matrix constructing module 203 is configured to form a two-dimensional fault probability matrix from results output by each of the nonlinear support vector machine models by using the number of the nonlinear support vector machine models as rows and the occurrence probability of the fault as columns.
The normalized coefficient determination module 204 is configured to determine a normalized coefficient for the probability of occurrence of a fault in the two-dimensional fault probability matrix.
The descending order ranking module 205 is configured to determine a fault probability value corresponding to each fault according to the normalization coefficient, and rank all the fault probability values in a descending order.
The fault determining module 206 is configured to determine that the two kinds of faults have high probability and are correspondingly diagnosed as "unrecognizable" when a difference between the first large fault probability value and the second large fault probability value after the descending order is smaller than a set threshold; and if the difference between the first large fault probability value and the second large fault probability value is larger than a set threshold value, determining the fault type corresponding to the first large fault probability value as the current fault state.
As a preferred embodiment of the present invention, the nuclear power system fault diagnosis system provided by the present invention may further include: the device comprises a historical operating data acquisition module, a calibration sampling module and a nonlinear support vector machine model training module.
The historical operation data acquisition module is used for acquiring historical operation data of the nuclear power system of the same reactor type.
And the simulation data determining module is used for simulating the historical operation data by adopting a full-range simulator to obtain simulation data of various single accident data and complex accidents.
And the calibration sampling module is used for performing calibration sampling on the simulation data to obtain the fault category in the same reactor type nuclear power system and the fault occurrence probability corresponding to the fault category.
And the nonlinear support vector machine model training module is used for taking the historical operating data, the fault category and the fault occurrence probability corresponding to the fault category as a training data sample pair, and training the nonlinear support vector machine model to obtain a trained nonlinear support vector machine model.
As another embodiment of the present invention, the nonlinear support vector machine model training module specifically includes: the device comprises a classification accuracy determining unit, a parameter optimizing unit and a nonlinear support vector machine model training unit.
The classification accuracy determining unit is used for inputting the training sample pair into the nonlinear support vector machine model to obtain the classification accuracy of the nonlinear support vector machine model.
The parameter optimizing unit is used for taking the classification accuracy as a fitness function and performing self-adaptive optimization on the parameters to be optimized by adopting a particle swarm algorithm; the parameters to be optimized are penalty factors and the width of a kernel function in the nonlinear support vector machine model.
And the nonlinear support vector machine model training unit is used for taking the optimized punishment factor and the width of the kernel function as a new punishment factor and a new width of the kernel function in the nonlinear support vector machine model to obtain the trained nonlinear support vector machine model.
Preferably, the parameter optimizing unit specifically includes: the system comprises an initial population fitness determining subunit, a global optimal particle fitness determining subunit and a parameter optimizing subunit.
The initial population fitness determining subunit is used for initializing the inertial weight, the cognitive learning factor and the social learning factor of the particle swarm algorithm and determining the fitness of the initialized initial population.
The global optimal particle fitness determining subunit is configured to perform random variation on the particle swarm by using a self-adaptive large-scale variation assurance algorithm in an iterative process of the particle swarm algorithm, so as to determine global optimal particles and fitness of the global optimal particles in each iterative process.
And the parameter optimizing subunit is used for finishing the self-adaptive optimization of the parameter to be optimized when the accumulated times that the fitness of the global optimal particle is greater than or equal to the set fitness threshold value is greater than or equal to the set value in the continuous iteration process.
As another embodiment of the present invention, the system further includes: and the normalization processing module is used for performing normalization processing on the operation data by using a dispersion normalization method.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. A method of diagnosing a fault in a nuclear power system, comprising:
acquiring a plurality of trained nonlinear support vector machine models which are the same as the number of subsystems to be diagnosed in the nuclear power system;
acquiring operation data of each subsystem to be diagnosed by using a sensor;
outputting the fault category of each subsystem to be diagnosed and the fault occurrence probability corresponding to the fault category by adopting the nonlinear support vector machine model according to the operation data;
taking the number of the nonlinear support vector machine models as a row and the occurrence probability of the faults as a column, and forming a two-dimensional fault probability matrix by the result output by each nonlinear support vector machine model;
determining a normalization coefficient for the fault occurrence probability in the two-dimensional fault probability matrix;
determining a fault probability value corresponding to each fault according to the normalization coefficient, and performing descending order arrangement on all fault probability values;
when the difference between the first large fault probability value and the second large fault probability value after descending order is smaller than a set threshold, the probability of occurrence of the two faults is determined to be large, and the two faults are correspondingly diagnosed as 'unrecognizable'; and if the difference between the first large fault probability value and the second large fault probability value is larger than a set threshold value, determining the fault type corresponding to the first large fault probability value as the current fault state.
2. The nuclear power system fault diagnosis method according to claim 1, wherein the training process of the non-linear support vector machine model includes:
acquiring historical operating data of the same reactor type nuclear power system;
simulating the historical operation data by adopting a full-range simulator to obtain simulation data of various single accident data and complex accidents;
carrying out calibration sampling on the simulation data to obtain the fault category in the same reactor type nuclear power system and the fault occurrence probability corresponding to the fault category;
and taking the historical operating data, the fault category and the fault occurrence probability corresponding to the fault category as a training data sample pair, and training the nonlinear support vector machine model to obtain a trained nonlinear support vector machine model.
3. The nuclear power system fault diagnosis method according to claim 2, wherein the training of the nonlinear support vector machine model by using the historical operating data, the fault category and the fault occurrence probability corresponding to the fault category as a training data sample pair specifically comprises:
inputting the training sample pair into the nonlinear support vector machine model to obtain the classification accuracy of the nonlinear support vector machine model;
taking the classification accuracy as a fitness function, and performing self-adaptive optimization on the parameters to be optimized by adopting a particle swarm algorithm; the parameters to be optimized are penalty factors and the width of a kernel function in the nonlinear support vector machine model;
and taking the optimized punishment factor and the width of the kernel function as a new punishment factor and the width of a new kernel function in the nonlinear support vector machine model to obtain the trained nonlinear support vector machine model.
4. The nuclear power system fault diagnosis method according to claim 3, wherein the classification accuracy is used as a fitness function, and a particle swarm optimization is adopted to perform adaptive optimization on the parameters to be optimized, and specifically comprises the following steps:
initializing the inertial weight, cognitive learning factors and social learning factors of the particle swarm algorithm, and determining the fitness of the initialized initial population;
in the iterative process of the particle swarm optimization, random variation is carried out on the particle swarm by adopting a self-adaptive large-scale variation guarantee algorithm so as to determine global optimal particles and the fitness of the global optimal particles in each iterative process;
in the continuous iteration process, when the accumulated times of the fitness of the global optimal particles is greater than or equal to the set fitness threshold value and greater than or equal to the set value, the self-adaptive optimization of the parameters to be optimized is completed.
5. The nuclear power system fault diagnosis method according to claim 1, further comprising, after the acquiring operation data of each subsystem to be diagnosed using the sensor,:
and normalizing the operation data by using a dispersion standardization method.
6. A nuclear power system fault diagnostic system, comprising:
the nonlinear support vector machine model acquisition module is used for acquiring a plurality of trained nonlinear support vector machine models which are the same as the number of subsystems to be diagnosed in the nuclear power system;
the operation data acquisition module is used for acquiring the operation data of each subsystem to be diagnosed by adopting a sensor;
the fault type and fault occurrence probability determining module is used for outputting the fault type in each subsystem to be diagnosed and the fault occurrence probability corresponding to the fault type by adopting the nonlinear support vector machine model according to the operation data;
the two-dimensional fault probability matrix construction module is used for forming a two-dimensional fault probability matrix by using the number of the nonlinear support vector machine models as rows and the fault occurrence probability as columns and outputting results of each nonlinear support vector machine model;
the normalization coefficient determining module is used for determining a normalization coefficient for the fault occurrence probability in the two-dimensional fault probability matrix;
the descending order ranking module is used for determining the fault probability value corresponding to each fault according to the normalization coefficient and carrying out descending order ranking on all the fault probability values;
a fault determining module, configured to determine that the two kinds of faults have high probability and are correspondingly diagnosed as "unrecognizable" when a difference between the first large fault probability value and the second large fault probability value after the descending order is smaller than a set threshold; and if the difference between the first large fault probability value and the second large fault probability value is larger than a set threshold value, determining the fault type corresponding to the first large fault probability value as the current fault state.
7. The nuclear power system fault diagnostic system of claim 6, further comprising:
the historical operating data acquisition module is used for acquiring historical operating data of the nuclear power system of the same reactor type;
the simulation data determining module is used for simulating the historical operation data by adopting a full-range simulator to obtain simulation data of various single accident data and complex accidents;
the calibration sampling module is used for performing calibration sampling on the simulation data to obtain the fault category in the same type of reactor type nuclear power system and the fault occurrence probability corresponding to the fault category;
and the nonlinear support vector machine model training module is used for taking the historical operating data, the fault category and the fault occurrence probability corresponding to the fault category as a training data sample pair, and training the nonlinear support vector machine model to obtain a trained nonlinear support vector machine model.
8. The nuclear power system fault diagnosis system of claim 7, wherein the non-linear support vector machine model training module specifically comprises:
a classification accuracy determining unit, configured to input the training sample pair into the nonlinear support vector machine model to obtain a classification accuracy of the nonlinear support vector machine model;
the parameter optimizing unit is used for taking the classification accuracy as a fitness function and carrying out self-adaptive optimization on the parameters to be optimized by adopting a particle swarm algorithm; the parameters to be optimized are penalty factors and the width of a kernel function in the nonlinear support vector machine model;
and the nonlinear support vector machine model training unit is used for taking the optimized punishment factor and the width of the kernel function as a new punishment factor and a new width of the kernel function in the nonlinear support vector machine model to obtain the trained nonlinear support vector machine model.
9. The nuclear power system fault diagnosis system according to claim 8, wherein the parameter optimizing unit specifically includes:
the initial population fitness determining subunit is used for initializing the inertial weight, the cognitive learning factor and the social learning factor of the particle swarm algorithm and determining the fitness of the initialized initial population;
the global optimal particle fitness determining subunit is used for performing random variation on the particle swarm by adopting a self-adaptive large-scale variation guarantee algorithm in the iterative process of the particle swarm algorithm so as to determine global optimal particles and the fitness of the global optimal particles in each iterative process;
and the parameter optimizing subunit is used for finishing the self-adaptive optimization of the parameter to be optimized when the accumulated times that the fitness of the global optimal particle is greater than or equal to the set fitness threshold value is greater than or equal to the set value in the continuous iteration process.
10. The nuclear power system fault diagnostic system of claim 6, further comprising:
and the normalization processing module is used for performing normalization processing on the operation data by using a dispersion normalization method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010654755.9A CN111767657B (en) | 2020-07-09 | 2020-07-09 | Nuclear power system fault diagnosis method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010654755.9A CN111767657B (en) | 2020-07-09 | 2020-07-09 | Nuclear power system fault diagnosis method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111767657A true CN111767657A (en) | 2020-10-13 |
| CN111767657B CN111767657B (en) | 2022-04-22 |
Family
ID=72725479
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010654755.9A Active CN111767657B (en) | 2020-07-09 | 2020-07-09 | Nuclear power system fault diagnosis method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111767657B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112326255A (en) * | 2020-11-02 | 2021-02-05 | 重庆大学 | Engine remote health monitoring system and monitoring method |
| CN114115195A (en) * | 2021-11-22 | 2022-03-01 | 江苏科技大学 | Fault diagnosis and fault-tolerant control method for underwater robot propeller |
| CN114323691A (en) * | 2021-12-28 | 2022-04-12 | 中国科学院工程热物理研究所 | Gas circuit fault diagnosis device and method for compressed air energy storage system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105955069A (en) * | 2016-06-12 | 2016-09-21 | 哈尔滨工程大学 | On-line-simulated-based nuclear power plant system level state monitoring method |
| CN107301884A (en) * | 2017-07-24 | 2017-10-27 | 哈尔滨工程大学 | A kind of hybrid nuclear power station method for diagnosing faults |
| CN107505133A (en) * | 2017-08-10 | 2017-12-22 | 滁州学院 | The probability intelligent diagnosing method of rolling bearing fault based on adaptive M RVM |
| CN110738274A (en) * | 2019-10-26 | 2020-01-31 | 哈尔滨工程大学 | nuclear power device fault diagnosis method based on data driving |
| CN110929768A (en) * | 2019-11-14 | 2020-03-27 | 国电大渡河检修安装有限公司 | Prediction method for machine fault |
| CN111240306A (en) * | 2020-04-26 | 2020-06-05 | 南京市产品质量监督检验院 | Self-adaptive distribution transformer fault diagnosis system and diagnosis method thereof |
-
2020
- 2020-07-09 CN CN202010654755.9A patent/CN111767657B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105955069A (en) * | 2016-06-12 | 2016-09-21 | 哈尔滨工程大学 | On-line-simulated-based nuclear power plant system level state monitoring method |
| CN107301884A (en) * | 2017-07-24 | 2017-10-27 | 哈尔滨工程大学 | A kind of hybrid nuclear power station method for diagnosing faults |
| CN107505133A (en) * | 2017-08-10 | 2017-12-22 | 滁州学院 | The probability intelligent diagnosing method of rolling bearing fault based on adaptive M RVM |
| CN110738274A (en) * | 2019-10-26 | 2020-01-31 | 哈尔滨工程大学 | nuclear power device fault diagnosis method based on data driving |
| CN110929768A (en) * | 2019-11-14 | 2020-03-27 | 国电大渡河检修安装有限公司 | Prediction method for machine fault |
| CN111240306A (en) * | 2020-04-26 | 2020-06-05 | 南京市产品质量监督检验院 | Self-adaptive distribution transformer fault diagnosis system and diagnosis method thereof |
Non-Patent Citations (4)
| Title |
|---|
| HANG WANG 等: "A hybrid fault diagnosis methodology with support vector machine and improved particle swarm optimization for nuclear power plants", 《ISA TRANSACTIONS》 * |
| 彭刚等: "基于改进多分类概率SVM模型的变压器故障诊断", 《机械与电子》 * |
| 戴海发等: "基于一类SVM的综合导航系统信息故障检测方法", 《中国惯性技术学报》 * |
| 王航: "模型驱动的核电站混合式故障诊断策略研究", 《中国博士学位论文全文数据库 工程科技II辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112326255A (en) * | 2020-11-02 | 2021-02-05 | 重庆大学 | Engine remote health monitoring system and monitoring method |
| CN114115195A (en) * | 2021-11-22 | 2022-03-01 | 江苏科技大学 | Fault diagnosis and fault-tolerant control method for underwater robot propeller |
| CN114115195B (en) * | 2021-11-22 | 2024-03-15 | 江苏科技大学 | Underwater robot propeller fault diagnosis and fault tolerance control method |
| CN114323691A (en) * | 2021-12-28 | 2022-04-12 | 中国科学院工程热物理研究所 | Gas circuit fault diagnosis device and method for compressed air energy storage system |
| CN114323691B (en) * | 2021-12-28 | 2023-06-23 | 中国科学院工程热物理研究所 | Air path fault diagnosis device and method for compressed air energy storage system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111767657B (en) | 2022-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111767657B (en) | Nuclear power system fault diagnosis method and system | |
| CN110441065B (en) | Gas turbine on-line detection method and device based on LSTM | |
| CN105117602B (en) | A kind of metering device running status method for early warning | |
| CN113255848B (en) | Water turbine cavitation sound signal identification method based on big data learning | |
| CN106355030B (en) | A kind of fault detection method based on analytic hierarchy process (AHP) and Nearest Neighbor with Weighted Voting Decision fusion | |
| CN108921230A (en) | Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network | |
| CN109685366A (en) | Equipment health state evaluation method based on mutation data | |
| CN106371427A (en) | Industrial process fault classification method based on analytic hierarchy process and fuzzy fusion | |
| CN111881627B (en) | Nuclear power plant fault diagnosis method and system | |
| CN111680875B (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
| CN111899905B (en) | Fault diagnosis method and system based on nuclear power device | |
| CN109710661A (en) | Based on Global Genetic Simulated Annealing Algorithm to the method for high-pressure heater state analysis | |
| CN111273125A (en) | RST-CNN-based power cable channel fault diagnosis method | |
| Zhou et al. | Structural health monitoring of offshore wind power structures based on genetic algorithm optimization and uncertain analytic hierarchy process | |
| CN109685136A (en) | A kind of high-pressure heater status data analysis method | |
| CN117556347A (en) | Power equipment fault prediction and health management method based on industrial big data | |
| CN112036496A (en) | Nuclear power device fault diagnosis method and system | |
| CN112036087A (en) | Multi-strategy fused nuclear power key equipment fault diagnosis method and system | |
| CN117032165A (en) | Industrial equipment fault diagnosis method | |
| CN109242008B (en) | Compound fault identification method under incomplete sample class condition | |
| CN115713027A (en) | Transformer state evaluation method, device and system | |
| CN114943281B (en) | Intelligent decision-making method and system for heat pipe cooling reactor | |
| CN116383747A (en) | Anomaly detection method for generating countermeasure network based on multi-time scale depth convolution | |
| CN115153549A (en) | BP neural network-based man-machine interaction interface cognitive load prediction method | |
| Najar et al. | Comparative Machine Learning Study for Estimating Peak Cladding Temperature in AP1000 Under LOFW |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |