CN112036480A - Ship refrigeration system fault diagnosis method and device and storage medium - Google Patents
Ship refrigeration system fault diagnosis method and device and storage medium Download PDFInfo
- Publication number
- CN112036480A CN112036480A CN202010890746.XA CN202010890746A CN112036480A CN 112036480 A CN112036480 A CN 112036480A CN 202010890746 A CN202010890746 A CN 202010890746A CN 112036480 A CN112036480 A CN 112036480A
- Authority
- CN
- China
- Prior art keywords
- data
- refrigeration system
- support vector
- fault
- vector machine
- 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.)
- Pending
Links
- 238000005057 refrigeration Methods 0.000 title claims abstract description 55
- 238000003745 diagnosis Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012706 support-vector machine Methods 0.000 claims abstract description 87
- 239000002245 particle Substances 0.000 claims abstract description 24
- 238000005457 optimization Methods 0.000 claims abstract description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 12
- 238000013507 mapping Methods 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 10
- 210000000349 chromosome Anatomy 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 238000000513 principal component analysis Methods 0.000 abstract description 19
- 230000035945 sensitivity Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 20
- 238000010606 normalization Methods 0.000 description 13
- 239000011159 matrix material Substances 0.000 description 9
- 238000012360 testing method Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 5
- 230000008878 coupling Effects 0.000 description 4
- 238000010168 coupling process Methods 0.000 description 4
- 238000005859 coupling reaction Methods 0.000 description 4
- 238000002405 diagnostic procedure Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 239000000498 cooling water Substances 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000006185 dispersion Substances 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012847 principal component analysis method Methods 0.000 description 2
- 239000003507 refrigerant Substances 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 238000002759 z-score normalization Methods 0.000 description 2
- 238000004378 air conditioning Methods 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000037213 diet Effects 0.000 description 1
- 235000005911 diet Nutrition 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 238000001704 evaporation Methods 0.000 description 1
- 230000008020 evaporation Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B63—SHIPS OR OTHER WATERBORNE VESSELS; RELATED EQUIPMENT
- B63J—AUXILIARIES ON VESSELS
- B63J2/00—Arrangements of ventilation, heating, cooling, or air-conditioning
- B63J2/12—Heating; Cooling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Abstract
The invention provides a fault diagnosis method and device for a ship refrigeration system and a storage medium. The method comprises the following steps: collecting data of a ship refrigeration system, wherein the data comprises system operation data and a working condition type corresponding to the operation data; the method comprises the steps of utilizing a trained fault recognition model to carry out fault recognition on data of the ship refrigeration system, wherein the fault recognition model is used for classifying system operation data to obtain a working condition type corresponding to the operation data, the fault recognition model is a support vector machine model, and penalty factors and nuclear parameters of the support vector machine model are obtained based on particle swarm optimization. The method for optimizing the support vector machine based on the principal component analysis and the particle swarm optimization carries out fault diagnosis on the ship refrigeration system, so that the fault diagnosis accuracy of the ship refrigeration system is improved, the diagnosis time is shortened, and the fault diagnosis sensitivity is improved.
Description
Technical Field
The invention relates to the technical field of ship turbine engineering, in particular to a fault diagnosis method and device for a ship refrigeration system and a storage medium.
Background
The refrigeration system is used as an important auxiliary device and can ensure the air conditioning and the diet refrigeration of the ship. The internal structure of the refrigerating system is complex, so that the operation condition of the refrigerating system is unstable, the freshness of food in the refrigerated cabinet is damaged, the living comfort of crews is reduced, and the service life of equipment is shortened.
The application of the fault diagnosis technology in the refrigeration system is relatively late compared with other fields, and the common fault diagnosis methods mainly comprise three types: methods based on quantitative physical models, methods based on qualitative physical models, and methods based on operational data. The diagnostic method based on the operation data does not need to depend on a system model, the operation data is analyzed and identified, the application range is wide, and the methods such as a neural network algorithm, a support vector machine and a probabilistic neural network become popular methods in the field.
Disclosure of Invention
The ship refrigeration system fault diagnosis method, device and storage medium are provided for solving the technical problem that a ship refrigeration system is complex in structure and difficult to diagnose faults. According to the method, the optimized support vector machine model is mainly utilized to carry out fault diagnosis on 11 working conditions of the ship refrigeration system, and the PCA is combined to carry out dimensionality reduction processing on the data, so that redundant, miscellaneous and invalid information is effectively removed, the correlation interference among the data is reduced, the diagnosis effect is ensured, and the diagnosis efficiency is improved.
The technical means adopted by the invention are as follows:
a method of fault diagnosis for a marine refrigeration system, comprising:
collecting data of a ship refrigeration system, wherein the data comprises system operation data and a working condition type corresponding to the operation data;
the method comprises the steps of utilizing a trained fault recognition model to carry out fault recognition on data of the ship refrigeration system, wherein the fault recognition model is used for classifying system operation data to obtain a working condition type corresponding to the operation data, the fault recognition model is a support vector machine model, and penalty factors and nuclear parameters of the support vector machine model are obtained based on particle swarm optimization.
Further, after the data of the ship refrigeration system is collected, the method further comprises the following steps:
and mapping the operation data into a new characteristic space after linear transformation to generate operation characteristic data, wherein the dimensionality of the operation characteristic data is lower than that of the operation data.
Further, after the data of the ship refrigeration system is collected, the method further comprises the following steps:
and carrying out standardization processing on the operation data, and mapping the operation data to a [0,1] interval to generate operation characteristic data.
Further, optimizing and acquiring a penalty factor and a kernel parameter of the support vector machine model based on a particle swarm optimization algorithm, wherein the method comprises the following steps:
forming individual chromosomes by the penalty factors and the radial basis kernel parameters of the support vector machine;
taking the identification accuracy of the support vector machine as an individual fitness function, wherein the identification accuracy is the fault identification accuracy of the support vector machine on historical data of a ship refrigeration system;
and optimizing the penalty factor and the nuclear parameter according to the fitness function based on a particle swarm algorithm, re-determining a classification judgment function of the support vector machine, and storing excellent individuals.
A marine vessel refrigeration system fault diagnostic apparatus comprising:
the data acquisition unit is used for acquiring data of the ship refrigeration system, and comprises system operation data and working condition types corresponding to the operation data;
the fault identification unit is used for carrying out fault identification on the data of the ship refrigeration system by using a trained fault identification model, the fault identification model is used for classifying the system operation data to obtain a working condition type corresponding to the operation data, the fault identification model is a support vector machine model, and penalty factors and nuclear parameters of the support vector machine model are obtained based on particle swarm optimization.
Furthermore, the data acquisition unit comprises a dimensionality reduction module, which is used for mapping the operation data into a new feature space after linear transformation to generate operation feature data, wherein the dimensionality of the operation feature data is lower than that of the operation data.
Further, the data acquisition unit comprises a standardization module, and the standardization module is used for carrying out standardization processing on the operation data, mapping the operation data into a [0,1] interval and generating operation characteristic data.
Further, the device also comprises a model construction unit, wherein the model construction unit is used for constructing the support vector machine model and optimizing the penalty factor and the nuclear parameter of the support vector machine, and the process comprises the following steps:
forming individual chromosomes by the penalty factors and the radial basis kernel parameters of the support vector machine;
taking the identification accuracy of the support vector machine as an individual fitness function, wherein the identification accuracy is the fault identification accuracy of the support vector machine on historical data of a ship refrigeration system;
and optimizing the penalty factor and the nuclear parameter according to the fitness function based on a particle swarm algorithm, re-determining a classification judgment function of the support vector machine, and storing excellent individuals.
A computer-readable storage medium having a set of computer instructions stored therein; the set of computer instructions, when executed by a processor, implement a vessel refrigeration system fault diagnosis method as in any one of the above.
Compared with the prior art, the invention has the following advantages:
1. the method selects the data characteristics by using a Principal Component Analysis (PCA) model, removes redundant data characteristics and correlation influence among data, and realizes the dimensionality reduction and decoupling of the data.
2. The invention establishes a multi-classification Support Vector Machine (SVM) model, and inputs data after dimension reduction processing into the SVM model for multi-working-condition fault model training.
3. The invention uses the global search capability of Particle Swarm Optimization (PSO), optimizes the penalty factor C and the kernel parameter sigma of the SVM according to the Particle fitness value, the individual extreme value and the size of the population extreme value, and re-determines the SVM classification decision function.
4. The invention can reserve the parameters of the model after being reserved, and can accelerate the convergence speed and precision of calculation when the PSO algorithm is called again to optimize the SVM.
The method for optimizing the support vector machine based on the principal component analysis and the particle swarm optimization carries out fault prediction and diagnosis on the ship refrigeration system, so that the fault diagnosis accuracy of the ship refrigeration system is improved, the diagnosis time is shortened, the fault diagnosis sensitivity is improved, a foundation is laid for green shipping, safe sailing, driving and protecting, unmanned ship development, and tiling is added for energy conservation and emission reduction.
Based on the reasons, the invention can be widely popularized in the field of ship turbine engineering.
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 description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flow chart of a fault diagnosis method for a refrigeration system of a ship according to the invention.
FIG. 2 is a pareto chart of principal component contribution ratios in an example.
FIG. 3 is a PSO-SVM fitness curve diagram of an initial fault in an embodiment.
FIG. 4 shows the predicted result of the PSO-SVM for the initial fault in the embodiment.
FIG. 5 is a PCA-PSO-SVM fitness curve diagram in the example.
FIG. 6 shows the PCA-PSO-SVM recognition result in the embodiment.
FIG. 7 shows the diagnosis accuracy of different models under 6 conditions of normal and late faults in the embodiment.
Detailed Description
In order to make the technical solutions of the present invention better understood, 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.
As shown in fig. 1, the present invention provides a method for diagnosing a fault of a refrigeration system of a ship, including: collecting data of a ship refrigeration system, wherein the data comprises system operation data and a working condition type corresponding to the operation data; the method comprises the steps of utilizing a trained fault recognition model to carry out fault recognition on data of the ship refrigeration system, wherein the fault recognition model is used for classifying system operation data to obtain a working condition type corresponding to the operation data, the fault recognition model is a support vector machine model, and penalty factors and nuclear parameters of the support vector machine model are obtained based on particle swarm optimization.
This embodiment uses boats and ships refrigerating system simulation experiment platform to carry out the fault experiment, has gathered 11 kinds of operating mode samples, specifically includes normal, 5 kinds of initial stage trouble operating modes and 5 middle and later stage trouble operating modes. During fault diagnosis of the ship refrigeration system, data acquisition comprises a plurality of observation point variables, so that a group of observation data is formed, and a plurality of groups of observation data in a certain time form an original data matrix. At present, the sampling time interval is set to be generally smaller during data acquisition, the acquired data volume is larger, and strong coupling exists between the data of the ship refrigeration system. Principal component analysis belongs to a feature extraction method of multivariate statistical regression, which is characterized in that data features in an original data matrix are subjected to linear transformation and are converted into a new feature space, basis vectors (principal components) of the new feature space can contain most of information of original variables, correlation influence is removed, and the dimension of the data matrix is reduced.
Further, in the principal component analysis, since z-score normalization processing is performed on the sample data in order to eliminate the dimensional influence of different characteristic variables, normalization processing is not required for the model using the principal component analysis. For the model without principal component analysis, the normalization method adopted herein is min-max normalization (also called dispersion normalization), and the sample data is mapped between [0,1], so that the operation is faster.
The SVM has the main idea that a classification hyperplane is established, and the isolation margin of positive and negative examples is maximized according to the structural risk minimization principle. The method has unique advantages in the recognition problems of few samples, nonlinearity, high dimensionality and the like, is one of the hot spots of the current machine learning research, and is particularly widely applied in the fields of pattern recognition, regression analysis, time series prediction and the like.
The punishment factor C and the nuclear parameter sigma of the SVM modeling method have direct influence on the recognition result and are influenced by the specific conditions of the punishment factor C and the nuclear parameter sigma: the model error tends to decrease first and then increase with the increase of C, and the fitting situation changes from the over-learning phenomenon to the under-learning phenomenon with the increase of sigma. Therefore, the method optimizes the penalty factor and the kernel parameter of the SVM by using the particle swarm optimization. The method specifically comprises the following steps:
(1) and initializing the population. Two parameters need to be determined for the SVM, namely a penalty factor C and a radial basis kernel parameter σ, which are formed into chromosomes of an individual.
(2) A fitness function. In order to establish an SVM model with a stable recognition effect, the recognition accuracy of the SVM is used as an individual fitness evaluation value.
(3) And (4) establishing and predicting the SVM model. And establishing a multi-classification SVM model by using comprehensive characteristic data subjected to dimensionality reduction according to a principal component analysis method as a data set. The standard SVM can only solve the problem of two classifications, 11 types of SVM are divided into 2 groups, one group comprises normal working conditions and 5 middle primary fault working conditions, the other group comprises normal working conditions and 5 later stage fault working conditions, a one-to-one multi-classification method is adopted, and each group is constructed
And the standard two-classification SVM is used for determining which working condition the SVM result finally belongs to by adopting a voting method. Before modeling, internal structure parameters of the SVM need to be determined, a kernel function type is determined, and a penalty factor and a smoothing factor are assigned.
And then optimizing the SVM based on the particle swarm algorithm, optimizing the penalty factor and the kernel parameter of the SVM according to the fitness function, re-determining the SVM classification judgment function, and storing excellent individuals so as to be used when a new PSO model is built, reduce the operation time and improve the calculation precision.
The solution of the invention is further illustrated by the following specific application examples.
In the embodiment, a certain refrigeration experiment system is adopted to carry out fault simulation experiment, 5 fault working conditions are graded according to the severity of the fault, each fault degree is graded into three grades, namely an initial fault, a middle fault and a later fault, wherein the later fault represents the most serious condition of the fault. The initial fault is small in wear coefficient, small in data expression deviation and large in fault identification difficulty coefficient, so that the method selects the initial fault experimental data for analysis, but comprehensively analyzes the initial fault and the later fault in the summary part.
1. Data source and preprocessing
(1) Extraction of sample data
The experimental system collects 6 working condition experimental samples including normal working conditions and initial faults, the specific working condition serial numbers and the working condition names are shown in table 1, the symbols and the names of 12 characteristic variables are shown in table 2, 200 groups of samples are collected in each working condition mode, the later fault working condition data samples are shown in table 3, the total number of the 1000 groups of experimental samples is 1000 x 12.
(2) Normalization of sample data
In principal component analysis, since z-score normalization processing is performed on sample data in order to eliminate the dimensional influence of different characteristic variables, normalization processing is not required for a model using principal component analysis. For the model without principal component analysis, the normalization method adopted herein is min-max normalization (also called dispersion normalization), and the sample data is mapped between [0,1], so that the operation is faster. The main form of the normalization function is as follows:
wherein x ismaxIs the maximum value, x, of this type of sample dataminIs the minimum value of the sample data of the class, x is the numerical value before normalization, x*Is normalized value with the value of [0,1]]In the meantime. The normalization process of the variables can be implemented with MATLAB's own function mapminmax, since the function defaults to normalization to [ -1,1]In between, the function needs to be specified to map the variable to [0,1]]An interval.
TABLE 1 working condition number and name
| Working condition number | Type of operating mode |
| y0 | Normal operating mode |
| y1 | Leakage of suction valve |
| y2 | Leakage of exhaust valve |
| y3 | Ice plug of expansion valve |
| y4 | Condenser dirt |
| y5 | The refrigerant containing non-condensable gas |
TABLE 2 characteristic variable symbols and names
| Serial number | Variable sign | Variable names | Serial number | Variable sign | Variable names |
| 1 | Ps | Compressor inlet pressure | 7 | Pe | Evaporation pressure of evaporator |
| 2 | Pd | Discharge pressure of compressor | 8 | Tra | Cabin air temperature |
| 3 | Tc | Condenser temperature | 9 | Hsa | Enthalpy of air supply |
| 4 | Ta | Temperature of air supply of blower | 10 | Hra | Enthalpy of cabin air |
| 5 | Tin | Inlet temperature of cooling water | 11 | Wsa | Humidity of air supply |
| 6 | Tout | Outlet temperature of cooling water | 12 | Wra | Humidity of cabin air |
TABLE 3 initial Fault Condition raw data sample
2. Principal component analysis and improved particle swarm optimization support vector machine
(1) Data dimension reduction
The acquired 12-dimensional data is reduced to 4 dimensions by using a principal component analysis method, and a principal component calculation and extraction formula is as follows:
Ti＝Xpi(i＝1，2，…，b) (2)
principal component TiRepresenting the data matrix X in the eigenvector piProjection in the direction, TiThe larger the size, the higher the value at piThe more information in the direction representing X. In order to reduce the number of principal components and to include the most information in the data matrix X, the number of remaining principal components is determined by the principal component contribution method (CPV).
Wherein, CPViIs the variance contribution, CPV, of the ith principal componentsThe cumulative variance contribution rate of the first k principal components is obtained, and the principal components with the contribution rate of more than 85% are generally selected.
From equation (2), the eigenvalue and variance contribution ratio shown in table 4 are calculated. Characteristic value lambdaiArranged from large to small, the columns corresponding to the numbers 1 to 12 correspond to the characteristic value lambdaiCharacteristic vector p ofi，CPViIs the variance contribution rate of the corresponding principal component.
TABLE 4 principal Components analysis results
| i | λi | CPVi/％ | 1 | 2 | 3 | 4 | … | 10 | 11 | 12 |
| 1 | 0.9444 | 75.4790 | -0.3702 | 0.1000 | -0.1042 | 0.0363 | … | -0.0468 | 0.0206 | -0.0001 |
| 2 | 0.1818 | 14.5276 | 0.0205 | 0.5410 | 0.2231 | 0.0436 | … | -0.0312 | -0.0040 | -0.0007 |
| 3 | 0.0368 | 2.9419 | 0.0329 | 0.1982 | 0.6096 | 0.3614 | … | -0.0710 | -0.0148 | -0.0069 |
| 4 | 0.0259 | 2.0701 | -0.3951 | 0.0445 | -0.0885 | 0.2558 | … | 0.2875 | 0.6163 | -0.4473 |
| 5 | 0.0233 | 1.8656 | -0.3477 | 0.1064 | 0.1281 | -0.3169 | … | -0.1151 | 0.4988 | 0.5547 |
| 6 | 0.0167 | 1.3353 | 0.4415 | -0.1193 | 0.2497 | -0.3099 | … | 0.6809 | 0.2413 | -0.0594 |
| 7 | 0.0091 | 0.7244 | 0.2991 | -0.0960 | -0.2111 | 0.4325 | … | 0.0886 | 0.2485 | 0.6319 |
| 8 | 0.0070 | 0.5634 | -0.2263 | -0.0665 | 0.5907 | -0.0262 | … | 0.1270 | -0.1557 | 0.1561 |
| 9 | 0.0049 | 0.3955 | 0.1021 | 0.0588 | 0.0209 | 0.5266 | … | 0.0156 | 0.0015 | -0.0354 |
| 10 | 0.0010 | 0.0833 | 0.2053 | 0.5397 | -0.1593 | 0.1289 | … | 0.1286 | 0.0610 | -0.0056 |
| 11 | 0.0001 | 0.0110 | 0.1261 | 0.5568 | -0.1540 | -0.3338 | … | -0.0107 | -0.0294 | 0.0070 |
| 12 | 3.5772E-05 | 0.0029 | -0.4247 | 0.1101 | -0.1954 | 0.1062 | … | 0.6257 | -0.4711 | 0.2524 |
As shown in fig. 1, the cumulative variance contribution rate of the first 4 principal components in table 4 has exceeded 95%, i.e., the first 4 principal component feature data has been able to reflect the information of the original data feature above 95%. In order to further verify the fault diagnosis effect after the principal component analysis, principal components with the accumulated contribution rates of 95%, 98%, more than 99% and 100% are selected for model training and testing, and the corresponding numbers of the principal components are respectively 4, 6, 8 and 12.
Therefore, 12 characteristics in the original data set X are converted into 12 new comprehensive characteristics which are independent of each other after principal component analysis, as shown in table 5, by combining the analysis in the above, only the first 4 columns of comprehensive characteristic data in table 5 are needed to carry out fault diagnosis, and the dimensionality of the original data characteristics is reduced.
TABLE 5 newly generated comprehensive characteristic quantities
(2) Establishing SVM model
The data set is divided into a training set and a test set according to a ratio of 9:1, namely 1080 groups of data (180 groups of data in each working condition) containing 6 working conditions are used as the training set, and 120 groups of data in total containing 20 groups of data in each working condition are used as the test set. And establishing a multi-classification SVM model by using an LIBSVM tool box in MATLAB, determining which working condition each SVM result finally belongs to by using a voting method, wherein the kernel function type is designated as a radial basis function, and the penalty factor and the smoothness factor are both designated as 1.
(3) PSO optimization SVM
According to the SVM modeling principle, the penalty factor C and the kernel parameter sigma have direct influence on the recognition result. The SVM is specifically influenced by a penalty factor C and a nuclear parameter sigma: the model error tends to decrease first and then increase with the increase of C, and the fitting situation changes from the over-learning phenomenon to the under-learning phenomenon with the increase of sigma. Therefore, a particle swarm algorithm is used herein to optimize the penalty factor and the kernel parameter of the SVM.
3. Numerical experiment and result analysis
(1) PSO-SVM algorithm evaluation
And (3) according to the normal working condition and 5 kinds of initial fault data, the data are as follows: the 1 proportion is divided into a training set and a testing set, the size of the training set is 900 × 12, the size of the testing set is 100 × 12, SVM and PSO-SVM models are trained and tested, the number of particle swarm clusters is 20, the iteration number is 100 generations, an acceleration constant C1 is 1.5, and an acceleration constant C2 is 1.7. Fitness curves and fault recognition effects of the penalty factors and the kernel parameters of the PSO optimization SVM are shown in the following fig. 3 and 4, and finally obtained optimal penalty factors and kernel parameters are respectively 100 and 4.7243.
(2) Evaluation by PCA-PSO-SVM algorithm
Data of the first 4 principal component features in 5 initial fault training sets and test sets in table 5 are extracted, the size of the new training set is 900 × 4, the size of the new test set is 100 × 4, a PSO-SVM model is retrained and tested, fitness curves and fault recognition effects of penalty factors and kernel parameters of a PSO optimization SVM are shown in fig. 5 and 6, and finally obtained optimal penalty factors and kernel parameters are 87.3472 and 16.5334 respectively.
(3)5 kinds of initial fault diagnosis result analysis
In order to verify the rationality of the principal component analysis and highlight the fault diagnosis effect after the principal component analysis, on the basis of only selecting 4 principal components, principal components with the accumulated contribution rates of 98%, more than 99% and 100% are selected to train and test a PSO-SVM model, the numbers of the corresponding principal components are respectively 6, 8 and 12, the principal components are respectively represented by PCA98-PSO-SVM, PCA99-PSO-SVM and PCA100-PSO-SVM, the initial fault is diagnosed, and the PSO-SVM model and the PCA-PSO-SVM model are compared in the diagnosis result, as shown in Table 6.
TABLE 6 diagnosis of initial faults by different models
| Diagnostic method | Overall diagnostic accuracy/%) | Algorithm running time/S |
| PCA-PSO-SVM | 91％ | 101 |
| PCA98-PSO- |
89％ | 106 |
| PCA99-PSO- |
88％ | 112 |
| PCA100-PSO- |
90％ | 119 |
| PSO- |
90％ | 122 |
In order to deeply analyze the fault false alarm (misdiagnosis between fault conditions), a confusion matrix of different model prediction results and real results is established as shown in table 7, wherein the larger the elements of the main diagonal line in the confusion matrix are, the smaller the elements at other positions are, and the better the recognition effect of the model is.
The comparison results of the initial fault diagnosis by the different models in tables 6 and 7 show that after the initial fault data is subjected to principal component analysis, the feature quantities corresponding to the first 4 principal components are adopted to train the PSO-SVM model, and the PCA-PSO-SVM model is optimal in the aspects of the diagnosis accuracy, the diagnosis time and the diagnosis confusion matrix of the initial fault, so that the accuracy of the pareto chart of the principal component contribution rate of the graph 2 is verified.
TABLE 7 confusion matrix of different models for initial fault diagnosis
TABLE 86 diagnosis results of different models under different working conditions
| Diagnostic method | Overall diagnostic accuracy/%) | Algorithm run time consuming |
| SVM | 84.16 | 0.34 |
| PCA-SVM | 83.34 | 0.2 |
| GA-SVM | 95.83 | 210.78 |
| PCA-GA-SVM | 95 | 239.53 |
| PSO-SVM | 97.5 | 166.8 |
| PCA-PSO-SVM | 96.7 | 100.5 |
(4) Normal working condition and 5 later-stage fault diagnosis results
The experimental results of different diagnostic methods under the normal working condition and 5 later-stage fault working conditions of the ship refrigeration system are shown in fig. 7 and table 8, the accuracy of the SVM model in identifying the faults of the normal working condition y0, suction valve leakage y1, exhaust valve leakage y2, expansion valve ice plug y3 and condenser dirt y4 is very high, but the accuracy of identifying the faults of refrigerant containing non-condensable gas y5 is very low, which directly leads to the integral identification accuracy of the SVM to be only 84.16%; the GA-SVM model ensures that the recognition accuracy of the former 5 working conditions is very high, and simultaneously, the recognition accuracy of the y5 working condition is improved to 85 percent, so that the overall accuracy is improved by more than 10 percent compared with that of the SVM model, but the recognition accuracy of the expansion valve ice plug y3 working condition is reduced compared with that of the SVM model, and the recognition accuracy of the y3 working condition in the PCA-GA-SVM model is only 80 percent; compared with the GA-SVM model, the PSO-SVM model successfully solves the problem that the recognition accuracy of the SVM model in the y5 working condition and the recognition accuracy of the GA-SVM model in the y3 working condition are low, and the diagnosis time is greatly shortened compared with the GA-SVM model, so that the effectiveness of the particle swarm optimization on the optimization of the support vector machine model is proved in the diagnosis of the normal working condition and 5 later-stage faults of the ship refrigeration system.
4. Conclusion
It can be seen from the comparative analysis of different models in fig. 7, table 6 and table 8 that, no matter the initial fault or the later fault is diagnosed, no matter the diagnosis of a single working condition or the overall diagnosis is performed, the PCA is used to perform the dimensionality reduction processing on the data set, and after the PSO is used to optimize the penalty factor C and the kernel parameter σ of the SVM, the diagnosis time of the model can be effectively shortened on the premise of ensuring the diagnosis accuracy.
Corresponding to the fault diagnosis method in the invention, the invention also discloses a fault diagnosis device for the ship refrigeration system, which comprises the following steps: the data acquisition unit is used for acquiring data of the ship refrigeration system, and comprises system operation data and working condition types corresponding to the operation data; the fault identification unit is used for carrying out fault identification on the data of the ship refrigeration system by using a trained fault identification model, the fault identification model is used for classifying the system operation data to obtain a working condition type corresponding to the operation data, the fault identification model is a support vector machine model, and penalty factors and nuclear parameters of the support vector machine model are obtained based on particle swarm optimization.
Furthermore, the data acquisition unit comprises a dimensionality reduction module, which is used for mapping the operation data into a new feature space after linear transformation to generate operation feature data, wherein the dimensionality of the operation feature data is lower than that of the operation data.
Further, the data acquisition unit comprises a standardization module, and the standardization module is used for carrying out standardization processing on the operation data, mapping the operation data into a [0,1] interval and generating operation characteristic data.
Further, the device also comprises a model construction unit, wherein the model construction unit is used for constructing the support vector machine model and optimizing the penalty factor and the nuclear parameter of the support vector machine, and the process comprises the following steps:
forming individual chromosomes by the penalty factors and the radial basis kernel parameters of the support vector machine;
taking the identification accuracy of the support vector machine as an individual fitness function, wherein the identification accuracy is the fault identification accuracy of the support vector machine on historical data of a ship refrigeration system;
and optimizing the penalty factor and the nuclear parameter according to the fitness function based on a particle swarm algorithm, re-determining a classification judgment function of the support vector machine, and storing excellent individuals.
A computer-readable storage medium having a set of computer instructions stored therein; the set of computer instructions, when executed by the processor, implement any of the ship refrigeration system fault diagnosis methods described above.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (9)
1. A method of fault diagnosis for a refrigeration system of a marine vessel, comprising:
collecting data of a ship refrigeration system, wherein the data comprises system operation data and a working condition type corresponding to the operation data;
the method comprises the steps of utilizing a trained fault recognition model to carry out fault recognition on data of the ship refrigeration system, wherein the fault recognition model is used for classifying system operation data to obtain a working condition type corresponding to the operation data, the fault recognition model is a support vector machine model, and penalty factors and nuclear parameters of the support vector machine model are obtained based on particle swarm optimization.
2. The method of claim 1, wherein after collecting the data of the marine refrigeration system, the method further comprises:
and mapping the operation data into a new characteristic space after linear transformation to generate operation characteristic data, wherein the dimensionality of the operation characteristic data is lower than that of the operation data.
3. The method of claim 1, wherein after collecting the data of the marine refrigeration system, the method further comprises:
and carrying out standardization processing on the operation data, and mapping the operation data to a [0,1] interval to generate operation characteristic data.
4. The method for fault diagnosis of a ship refrigeration system according to claim 2, wherein the step of obtaining the penalty factor and the nuclear parameter of the support vector machine model based on particle swarm optimization comprises the steps of:
forming individual chromosomes by the penalty factors and the radial basis kernel parameters of the support vector machine;
taking the identification accuracy of the support vector machine as an individual fitness function, wherein the identification accuracy is the fault identification accuracy of the support vector machine on historical data of a ship refrigeration system;
and optimizing the penalty factor and the nuclear parameter according to the fitness function based on a particle swarm algorithm, re-determining a classification judgment function of the support vector machine, and storing excellent individuals.
5. A ship refrigeration system fault diagnosis device, comprising:
the data acquisition unit is used for acquiring data of the ship refrigeration system, and comprises system operation data and working condition types corresponding to the operation data;
the fault identification unit is used for carrying out fault identification on the data of the ship refrigeration system by using a trained fault identification model, the fault identification model is used for classifying the system operation data to obtain a working condition type corresponding to the operation data, the fault identification model is a support vector machine model, and penalty factors and nuclear parameters of the support vector machine model are obtained based on particle swarm optimization.
6. The fault diagnosis device for the ship refrigeration system as claimed in claim 5, wherein the data acquisition unit comprises a dimensionality reduction module for mapping the operation data into a new feature space after linear transformation to generate operation feature data, and the dimensionality of the operation feature data is lower than that of the operation data.
7. The fault diagnosis device for the ship refrigeration system according to claim 5, wherein the data acquisition unit comprises a standardization module for standardizing the operation data, mapping the operation data into a [0,1] interval, and generating operation characteristic data.
8. The apparatus according to claim 6, wherein the apparatus further comprises a model construction unit for constructing a support vector machine model and optimizing penalty factors and nuclear parameters of the support vector machine, and the process comprises:
forming individual chromosomes by the penalty factors and the radial basis kernel parameters of the support vector machine;
taking the identification accuracy of the support vector machine as an individual fitness function, wherein the identification accuracy is the fault identification accuracy of the support vector machine on historical data of a ship refrigeration system;
and optimizing the penalty factor and the nuclear parameter according to the fitness function based on a particle swarm algorithm, re-determining a classification judgment function of the support vector machine, and storing excellent individuals.
9. A computer-readable storage medium having a set of computer instructions stored therein; the set of computer instructions, when executed by a processor, implement the ship refrigeration system fault diagnosis method of any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010890746.XA CN112036480A (en) | 2020-08-29 | 2020-08-29 | Ship refrigeration system fault diagnosis method and device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010890746.XA CN112036480A (en) | 2020-08-29 | 2020-08-29 | Ship refrigeration system fault diagnosis method and device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112036480A true CN112036480A (en) | 2020-12-04 |
Family
ID=73586315
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010890746.XA Pending CN112036480A (en) | 2020-08-29 | 2020-08-29 | Ship refrigeration system fault diagnosis method and device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112036480A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112597658A (en) * | 2020-12-28 | 2021-04-02 | 哈尔滨工程大学 | Multi-model fault diagnosis method for marine diesel engine based on working condition identification |
| CN113255965A (en) * | 2021-04-26 | 2021-08-13 | 大连海事大学 | Intelligent processing system for prognosis of degradation fault of radar transmitter |
| CN113361016A (en) * | 2021-06-30 | 2021-09-07 | 大连海事大学 | Ship auxiliary boiler fault diagnosis method and device |
| CN117076915A (en) * | 2023-10-17 | 2023-11-17 | 中海油能源发展股份有限公司采油服务分公司 | Intelligent fault attribution analysis method and system for FPSO crude oil process system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101021724A (en) * | 2006-06-19 | 2007-08-22 | 青岛鑫三利冷箱技术有限公司 | Cold-storage container micro controller fault diagnosing system |
| CN104021238A (en) * | 2014-03-25 | 2014-09-03 | 重庆邮电大学 | Lead-acid power battery system fault diagnosis method |
| WO2017128455A1 (en) * | 2016-01-25 | 2017-08-03 | 合肥工业大学 | Analogue circuit fault diagnosis method based on generalized multiple kernel learning-support vector machine |
| CN111177974A (en) * | 2019-12-24 | 2020-05-19 | 北京航空航天大学 | Structure small failure probability calculation method based on double-layer nested optimization and subset simulation |
| CN111291783A (en) * | 2020-01-15 | 2020-06-16 | 北京市燃气集团有限责任公司 | Intelligent fault diagnosis method, system, terminal and storage medium for gas pressure regulating equipment |
-
2020
- 2020-08-29 CN CN202010890746.XA patent/CN112036480A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101021724A (en) * | 2006-06-19 | 2007-08-22 | 青岛鑫三利冷箱技术有限公司 | Cold-storage container micro controller fault diagnosing system |
| CN104021238A (en) * | 2014-03-25 | 2014-09-03 | 重庆邮电大学 | Lead-acid power battery system fault diagnosis method |
| WO2017128455A1 (en) * | 2016-01-25 | 2017-08-03 | 合肥工业大学 | Analogue circuit fault diagnosis method based on generalized multiple kernel learning-support vector machine |
| CN111177974A (en) * | 2019-12-24 | 2020-05-19 | 北京航空航天大学 | Structure small failure probability calculation method based on double-layer nested optimization and subset simulation |
| CN111291783A (en) * | 2020-01-15 | 2020-06-16 | 北京市燃气集团有限责任公司 | Intelligent fault diagnosis method, system, terminal and storage medium for gas pressure regulating equipment |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112597658A (en) * | 2020-12-28 | 2021-04-02 | 哈尔滨工程大学 | Multi-model fault diagnosis method for marine diesel engine based on working condition identification |
| CN112597658B (en) * | 2020-12-28 | 2022-02-18 | 哈尔滨工程大学 | Multi-model fault diagnosis method for marine diesel engine based on working condition identification |
| CN113255965A (en) * | 2021-04-26 | 2021-08-13 | 大连海事大学 | Intelligent processing system for prognosis of degradation fault of radar transmitter |
| CN113361016A (en) * | 2021-06-30 | 2021-09-07 | 大连海事大学 | Ship auxiliary boiler fault diagnosis method and device |
| CN117076915A (en) * | 2023-10-17 | 2023-11-17 | 中海油能源发展股份有限公司采油服务分公司 | Intelligent fault attribution analysis method and system for FPSO crude oil process system |
| CN117076915B (en) * | 2023-10-17 | 2024-01-09 | 中海油能源发展股份有限公司采油服务分公司 | Intelligent fault attribution analysis method and system for FPSO crude oil process system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112036480A (en) | Ship refrigeration system fault diagnosis method and device and storage medium | |
| CN111999088A (en) | Ship refrigeration system fault diagnosis method and device and storage medium | |
| Li et al. | An improved decision tree-based fault diagnosis method for practical variable refrigerant flow system using virtual sensor-based fault indicators | |
| Li et al. | A semi-supervised approach to fault detection and diagnosis for building HVAC systems based on the modified generative adversarial network | |
| CN107844799B (en) | Water chilling unit fault diagnosis method of integrated SVM (support vector machine) mechanism | |
| CN110579709B (en) | Fault diagnosis method for proton exchange membrane fuel cell for tramcar | |
| CN113177594B (en) | Air conditioner fault diagnosis method based on Bayesian optimization PCA-extreme random tree | |
| CN114484731A (en) | Method and device for diagnosing faults of central air conditioner based on stacking fusion algorithm | |
| Li et al. | Diagnosis for multiple faults of chiller using ELM-KNN model enhanced by multi-label learning and specific feature combinations | |
| Sun et al. | Optimization of support vector regression model based on outlier detection methods for predicting electricity consumption of a public building WSHP system | |
| Liu et al. | Sensor fault detection and diagnosis method for AHU using 1-D CNN and clustering analysis | |
| CN113203589A (en) | Distributed fault diagnosis method and system for multi-split air conditioning system | |
| CN110795690A (en) | Wind power plant operation abnormal data detection method | |
| CN111723925A (en) | Method, device, equipment and medium for fault diagnosis of on-road intelligent train air conditioning unit | |
| Li et al. | Machine learning based diagnosis strategy for refrigerant charge amount malfunction of variable refrigerant flow system | |
| Zhou et al. | An online compressor liquid floodback fault diagnosis method for variable refrigerant flow air conditioning system | |
| Zeng et al. | A hybrid deep forest approach for outlier detection and fault diagnosis of variable refrigerant flow system | |
| CN114154254A (en) | Fault diagnosis method for electric actuator of gas turbine | |
| CN113469252A (en) | Extra-high voltage converter valve operation state evaluation method considering unbalanced samples | |
| Dey et al. | Unsupervised learning techniques for HVAC terminal unit behaviour analysis | |
| Li et al. | Improved convolutional neural network chiller early fault diagnosis by gradient-based feature-level model interpretation and feature learning | |
| CN115017978A (en) | Fault classification method based on weighted probability neural network | |
| CN116150687A (en) | Fluid pipeline leakage identification method based on multi-classification G-WLSTSVM model | |
| CN113051530A (en) | KDE-FA-based cold water unit fault feature characterization method | |
| CN112036479A (en) | Ship air conditioning system fault identification method and device and storage medium |
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 |