CN109540562A - A kind of water cooler method for diagnosing faults - Google Patents
A kind of water cooler method for diagnosing faults Download PDFInfo
- Publication number
- CN109540562A CN109540562A CN201811516128.8A CN201811516128A CN109540562A CN 109540562 A CN109540562 A CN 109540562A CN 201811516128 A CN201811516128 A CN 201811516128A CN 109540562 A CN109540562 A CN 109540562A
- Authority
- CN
- China
- Prior art keywords
- data
- screw
- sample
- sampling
- fault simulation
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M99/00—Subject matter not provided for in other groups of this subclass
- G01M99/005—Testing of complete machines, e.g. washing-machines or mobile phones
Abstract
A kind of water cooler method for diagnosing faults according to the present invention, comprising the following steps: S1 carries out fault simulation experiment to screw-type water chiller, centrifugal refrigerating machines and acquires screw-type water chiller, centrifugal refrigerating machines fault simulation experimental data；S2, to screw-type water chiller screw-type water chiller fault simulation experimental data, random selection forms the second training set data and test set data after steady state process and removal noise processed；S3 establishes original training set data；S4 obtains the over-sampling training sample set data of different over-sampling multiplying powers after carrying out over-sampling processing to original training set data；S5 is trained over-sampling training sample set data using SVM model, obtains diagnostic model；Test set is put into diagnostic model and is tested by S6, obtains the diagnostic result of different water cooler fault models.Diagnostic method of the invention can carry out fault diagnosis to different types of water cooler, have broad application prospects.
Description
Technical field
The invention belongs to refrigerating fields, and in particular to a kind of water cooler failure based on PCA-SMOTE-SVM algorithm is examined
Disconnected method.
Background technique
In existing water cooler fault diagnosis research, water cooler and its fault diagnosis model are corresponded, and one
Fault model can only diagnose a water cooler, when diagnosing different types of water cooler, need again to carry out new unit
A large amount of fault simulation experiment, the new diagnostic model of training all expend a large amount of resource and time, are unfavorable for water cooler failure
The popularization of diagnostic techniques in practical applications.
When there is a kind of fault diagnosis model of efficient water cooler class type, other less type cold water are being obtained
Unit malfunction test sample forms new training sample set, the at this time training of archetype when new samples are mixed with original sample
Collection and novel water cooler fault sample collection quantity difference are larger, form uneven sample, can make raw diagnostic model pair
Novel water cooler performance of fault diagnosis decline.
Summary of the invention
In order to solve the problems, such as uneven sample set, the present invention provides a kind of base using synthesis minority class oversampling technique
It is used to carry out failure to different types of water cooler in the water cooler method for diagnosing faults of PCA-SMOTE-SVM algorithm to examine
It is disconnected.
The present invention provides a kind of water cooler method for diagnosing faults, have the feature that, comprising the following steps:
S1 carries out fault simulation experiment to screw-type water chiller and acquires screw-type water chiller fault simulation experiment number
According to screw-type water chiller fault simulation experimental data includes normal data and fault simulation data, to centrifugal refrigerating machines
Fault simulation experiment is carried out, forms the first training set data after acquiring centrifugal refrigerating machines fault simulation experimental data, is centrifuged
Formula water cooler fault simulation experimental data includes normal data and fault simulation data；
S2, it is random after steady state process and removal noise processed to screw-type water chiller fault simulation experimental data
Selection the second training set data of composition and test set data；
S3 establishes original training set data, which includes the first training set data and the second training set
Data；
S4 after carrying out over-sampling processing to original training set data, obtains the over-sampling training sample of different over-sampling multiplying powers
This collection data；
S5 is trained over-sampling training sample set data using SVM model, obtains diagnostic model；
Test set is put into diagnostic model and is tested by S6, obtains the diagnostic result of different water cooler fault models.
In water cooler method for diagnosing faults provided by the invention, which is characterized in that further comprising the steps of:
S7 assesses diagnostic model using different evaluation methods, obtains optimal water cooler imbalance sampling
When excellent diagnostics model.
In addition, can also have the following features: wherein in water cooler method for diagnosing faults provided by the invention,
Screw cooling-water machine normal data includes group evaporator leaving water temperature data, condenser inflow temperature data, spool position data
And frequency data, screw-type water chiller fault simulation data include refrigerant leak data, condenser water flow deficiency data
Data are overcharged with refrigerant.
In addition, can also have the following features: wherein in water cooler method for diagnosing faults provided by the invention,
First training set data includes normal condition, leakage of refrigerant, condenser fouling, condenser water flow is insufficient, condenser is not containing
Solidifying property gas, evaporator water flow are insufficient, refrigerant overcharges and excessive each 1000 sample datas of lubricating oil.
In addition, can also have the following features: wherein in water cooler method for diagnosing faults provided by the invention,
The expression formula of SVM model are as follows:
In formula, sgn () is sign function, K (x, xi) it is to select RBF function, i.e.,
In formula: σ is the width of radial basis function, and σ is smaller, and the width of radial basis function is smaller, more selective,It is radial base nuclear parameter, g is bigger, and radial basis function is more selective.X is to differentiate sample；xiFor training sample, i is
Number of samples,For the solution of dual problem, b*For threshold value, N is sample total.
The action and effect of invention
Related water cooler method for diagnosing faults according to the present invention is provided using synthesis minority class oversampling technique
A kind of water cooler method for diagnosing faults based on PCA-SMOTE-SVM algorithm, can to different types of water cooler into
Row fault diagnosis has broad application prospects in the practical application of water cooler fault diagnosis technology.
Detailed description of the invention
Fig. 1 is to apply the signal in water cooler training to shine using SMOTE oversampling technique in the embodiment of the present invention
Piece；
Fig. 2 is the centrifugal refrigerating machines and screw-type water chiller training sample after normalizing in the embodiment of the present invention
The schematic three dimensional views of collection；
Fig. 3 is centrifugal refrigerating machines Yu the screw-type water chiller training of over-sampling 100% in the embodiment of the present invention
Sample set schematic three dimensional views；
Diagnostic model holistic diagnosis performance change schematic diagram when Fig. 4 is different over-sampling multiplying powers in the embodiment of the present invention；
Fig. 5 is all kinds of failure Sensitivity (recall ratio) values under different over-sampling multiplying powers in the embodiment of the present invention；
Fig. 6 is all kinds of failure Precision (precision ratio) values under different over-sampling multiplying powers in the embodiment of the present invention；And
Fig. 7 is all kinds of failure F-measure values under different over-sampling multiplying powers in the embodiment of the present invention.
Specific embodiment
It is real below in order to be easy to understand the technical means, the creative features, the aims and the efficiencies achieved by the present invention
Example combination attached drawing is applied to be specifically addressed water cooler method for diagnosing faults of the invention.
Embodiment
The present invention provides a kind of cold water based on PCA-SMOTE-SVM algorithm using synthesis minority class oversampling technique
Unit method for diagnosing faults is used to carry out fault diagnosis to different types of water cooler.
Fault simulation experiment is carried out to screw-type water chiller and acquires screw-type water chiller fault simulation experimental data,
Screw-type water chiller fault simulation experimental data includes normal data and fault simulation data.
Fault simulation experiment is carried out to centrifugal refrigerating machines, after acquiring centrifugal refrigerating machines fault simulation experimental data
The first training set data is formed, centrifugal refrigerating machines fault simulation experimental data includes normal data and fault simulation data.
Fault simulation experiment.Screw-type water chiller fault simulation tests the frequency conversion screw using 200 standard tons
Water cooler, fault simulation experiment be divided into normal data acquisition and fault simulation data acquisition, the duty parameter used for
Evaporator leaving water temperature, condenser inflow temperature, spool position and frequency.Main analog failure is refrigerant leakage, condenser
Water flow deficiency and refrigerant overcharge three kinds of failures.Centrifugal refrigerating machines fault data is ASHRAE1043-RP data, composition
Training set 1, comprising normal condition, leakage of refrigerant, condenser fouling, condenser water flow is insufficient, condenser gas containing incoagulability
Body, evaporator water flow are insufficient, refrigerant overcharges and excessive each 1000 sample datas of lubricating oil.
1) fault type
This experimental data is by centrifugal refrigerating machines -1043RP and frequency conversion screw water cooler fault simulation experimental data group
At.Troubleshooting type and initialism are shown in Table 1, and screw water cooling machine set only simulates condenser water flow deficiency, refrigerant
Leakage/charging amount is insufficient, refrigerant charging excess three classes failure, and band * is shown in table.
1 seven kinds of typical faults of table
Screw-type water chiller fault simulation experimental data is selected at random after steady state process and removal noise processed
Select the second training set data of composition and test set data.
Data prediction.Screw-type water chiller fault simulation experimental data need to be by steady state process, removal noise processed
Random selection composition training set 2 and test set afterwards.It include 100 Normal sample datas, 100 RefLeak samples in training set 2
Notebook data, 100 ReduCF sample datas and 100 RefOver sample datas；Test set includes 1000 Normal sample numbers
According to, 1000 RefLeak sample datas, 1000 ReduCF sample datas and 1000 RefOver sample datas.
Original training set data is established, which includes the first training set data and the second training set number
According to.
After carrying out over-sampling processing to original training set data, the over-sampling training sample set of different over-sampling multiplying powers is obtained
Data.
Using the original training set data of PCA-SMOTE technical treatment.Original training set data is by 8000 centrifugal cold water
Unit sample and 400 screw-type water chiller sample compositions, sample imbalance rate are 5%.
2) oversampling technique: PCA-SMOTE
SMOTE is significantly improved the over-fitting situation as caused by non-heuristic random oversampler method.The side SMOTE
The core concept of method is exactly that the new samples generated at random are inserted between minority class sample and its neighbour's sample, and doing so can increase
Add the number of minority class sample and then improves the class imbalance distribution of data set.The main step of PCA-SMOTE technology in embodiment
It is rapid as follows:
1. carrying out dimension-reduction treatment, pivot tribute to selected 17 characteristic parameters (being shown in Table 1) first with principle component analysis (PCA)
The rate of offering is selected as 99%.
1 characteristic parameter table of table
2. the unbalance factor for calculating centrifugal refrigerating machines sample set and screw-type water chiller sample set is 5%, design
Over-sampling multiplying power is 10%, 15%, 20% and 25%, finds out N number of similar another neighbouring sample to screw-type water chiller sample set
This (N=100,800,1200 and 1600).
The original training set sample of table 2 composition
3. synthesizing new sample referring to formula (1) to screw-type water chiller sample set.
S2OSi=S2i+r and(S2ij-Si) (1)
Wherein, S2iFor screw-type water chiller original sample collection；S2ijIndicate S2iJ-th adjacent to sample, j=1 ...,
N；S2OSiScrew-type water chiller sample set after indicating over-sampling；Rand indicates a random number between 0 to 1.It will be newborn
At screw-type water chiller fault sample be merged into centrifugal refrigerating machines sample set, the cooling-water machine after forming new balance
Group training set.
In order to show PCA-SMOTE treated sample set on two-dimensional surface, PCA pivot contribution rate is 99%.
Centrifugal refrigerating machines sample set and screw-type water chiller sample set carry out SMOTE over-sampling 400% after PCA is handled
Processing, be then normalized using MinMaxScaler, by formula (2) processing map.
Wherein, SiFor original sample；SminFor smallest sample in sample set；SmaxFor maximum sample in sample set；SstdFor warp
Sample after crossing normalization.
The uneven ratio for calculating centrifugal refrigerating machines and screw-type water chiller is 5% (8000:400), designed and adopts
Sample rate is 100%, 200%, 300% and 400% (being shown in Table 3).For the sample group of screw-type water chiller, it was found that N number of class
As adjacent sample (N=400,800,1200 and 1600).
Sample composition under the different over-sampling multiplying powers of table 3.
As shown in Figure 1, original training set OG is by 8000 centrifugal refrigerating machines samples and 400 screw water cooling machine set samples
This composition, unbalance factor 5%.Over-sampling, each over-sampling are carried out to S2 sample set using PCA-SMOTE oversampling technique
100%, sample 400.So that S2 sample set increases, S2-OS100 sample number is 800, S2-OS200 sample number is 1200
A, S2-OS300 sample number is 1600, S2-OS400 sample number is 2000.Unbalance factor is respectively 10%, 15%, 20%
With 25%.
Into PCA-SMOTE oversampling technique, when over-sampling multiplying power is 100%, 1 sample number of training set is constant；Training
Every class sample increases by 100 in collection 2, increases by 400 in total, forms training sample set when Oversampling100%, this
When training set sample imbalance rate be 10%.Training sample is carried out at over-sampling using PCA-SMOTE technology according to this rule
Reason, this multiplying power of over-sampling are 200~400%.
Fig. 2 is the centrifugal refrigerating machines and screw-type water chiller training sample after normalizing in the embodiment of the present invention
The schematic three dimensional views of collection, Centrifugal is centrifugal refrigerating machines in figure, and Screw is screw-type water chiller, and Normal is
Normal condition, RefLeak are leakage of refrigerant failure, and ReduCF condenser water flow deficiency failure, RefOver is refrigerant mistake
Fill failure.
Fig. 3 is centrifugal refrigerating machines Yu the screw-type water chiller training of over-sampling 100% in the embodiment of the present invention
Sample set schematic three dimensional views, Centrifugal is centrifugal refrigerating machines in figure, and Screw is screw-type water chiller, Normal
For normal condition, RefLeak is leakage of refrigerant failure, and ReduCF condenser water flow deficiency failure, RefOver is refrigerant
Overcharge failure.
As shown in Figure 2,3, centrifugal refrigerating machines sample is all more concentrated, and screw-type water chiller sample distribution is wider；
As can be seen that all more concentrating after Normal sample set over-sampling after PCA-SMOTE is handled, sample plyability is high；
RefLeak and RefOver sample set is more in partial region over-sampling；ReduCF original sample collection it is wide be distributed in space
Increase after carrying out PCA-SMOTE oversampling technique in multiple regions sample set in domain.
Over-sampling training sample set data are trained using SVM model, utilize cross validation and trellis search method
Optimal radial base nuclear parameter g in SVM model is found, diagnostic model is obtained.
Training SVM model.SVM is that a kind of foundation is new on the basis of Statistical Learning Theory and structural risk minimization
Type Learning machine, it seeks best compromise between the complexity and learning ability of model according to finite sample information, to obtain
Best Generalization Ability.SVM model uses optimal decision function formula (3)
In formula, sgn () is sign function, K (x, xi) it is to select RBF function, i.e.,
In formula: σ is the width of radial basis function, and σ is smaller, and the width of radial basis function is smaller, more selective.
It is radial base nuclear parameter, g is bigger, and radial basis function is more selective, and x is to differentiate sample；xiFor training sample, i is sample number
Mesh,For the solution of dual problem, b*For threshold value, N is sample total.
It is trained by the training set that PCA-SMOTE oversampling technique is handled well using SVM model, SVM model utilizes
Cross validation and trellis search method find radial base nuclear parameter g, to train excellent diagnostics model.
Test set is put into diagnostic model and is tested by water cooler fault diagnosis, obtains different water cooler failure moulds
The diagnostic result of type.
Diagnostic model is assessed using different evaluation methods, is obtained when optimal water cooler imbalance samples
Excellent diagnostics model.
Water cooler fault diagnosis.Test set is put into trained model to be tested, the diagnosis of different models is obtained
As a result, being assessed using different evaluation methods diagnostic model.Whole accuracy, G-mean value evaluate the whole of diagnostic model
Body performance；The diagnosis situation of Precision value, Sensitivity value and F-measure value evaluation all kinds of failures of water cooler.
The performance for comparing different over-sampling models obtains the diagnostic model when sampling of optimal water cooler imbalance, in reality
In promoted.
Evaluation criterion
The confusion matrix of the more classification diagnosis of water cooler is established according to two classification confusion matrixes, more classification confusion matrixes include
The information of the reality and prediction classification done about model.The property of this diagnostic model is assessed usually using the data in matrix
Energy.Table 4 shows more classification confusion matrix information.TA, TB, TC, TD are the correct classification samples of diagnostic model；FBA is normal shape
State is diagnosed as refrigerant leakage failure, belongs to false-alarm；FAB is that refrigerant leakage failure is diagnosed as normal condition, belongs to and fails to report；
FCB is that refrigerant leakage failure is diagnosed as condenser water flow deficiency failure, belongs to wrong report.Confusion matrix other information is with such
It pushes away.
More than 4 classification confusion matrixes of table
The present embodiment uses accuracy (Accuracy), G-mean value, precision ratio (Precision), recall ratio
(Sensitivity) and F-measure value evaluates the diagnosis performance of each class model.Table 5 shows each evaluation criterion formula.
Precision-A, Sensitivity-A and F-measure-A respectively represent the precision ratio, recall ratio and F- of normal condition
Measure value.What G-mean value indicated is the ensemble average value of minority class nicety of grading majority class nicety of grading, Lai Hengliang data
Collect whole classification performance；And F-measure index is a kind of evaluation of classification index for comprehensively considering recall ratio and precision ratio.
Each evaluation method formula of table 5
3) holistic diagnosis standard
Ma Xiusi related coefficient (the The Matthews Correlation being introduced into evaluation overall performance in two classification
Coefficient MCC), MCC returns to a value in [- 1 ,+1] range, classifies better closer to+1 representative model, calculates
Formula is as follows:
As can be drawn from Table 6, the training accuracy of five kinds of models is all 94% or so, with the difference of test accuracy compared with
Small, the smallest is PCA-SMOTE300%-SVM model, difference 1.39%.It is up to PCA-SMOTE100%-SVM model,
Only 2.48%, the generalization ability of the provable SVM model handled by PCA-SMOTE oversampling technique is still preferable.
Test accuracy it is highest be PCA-SMOTE100%-SVM model, from SVM model accuracy 95.90% promoted to
96.70%；The test accuracy of 3 kinds of models of over-sampling 200%, 300%, 400% has promotion, it is seen that oversampling technique energy
Enough make to keep higher rate of correct diagnosis in the lower situation of training sample balanced ratio.
For G-mean value, PCA-SMOTE100%-SVM model is put up the best performance, and being worth is 93.47%, is promoted compared with SVM model
1.57%.What G-mean value indicated is the ensemble average value of minority class nicety of grading majority class nicety of grading, Lai Hengliang data set
Whole classification performance.It can be seen that the holistic diagnosis performance of PCA-SMOTE-SVM model is preferable, compared with SVM model, performance is mentioned
It rises.
Putting up the best performance in MCC is PCA-SMOTE100%-SVM model, is 0.8776, is in all models closest to 1
, it is seen that PCA-SMOTE100%-SVM has optimal diagnosis performance.
To sum up, fault diagnosis is carried out to two type water coolers using the SVM diagnostic model of PCA-SMOTE oversampling technique
When, superior performance is still maintained, and on original SVM diagnostic model, performance is promoted, and PCA-SMOTE-SVM model is general
Change ability is stronger, and model generalization is preferable, and SMOTE oversampling technique can be good at using in the fault diagnosis of water cooler
Come.
The holistic diagnosis performance full edition of 6 five kinds of models of table
As shown in figure 4, the training accuracy of original SVM model, test accuracy, G-mean and MCC are respectively
93.62%, 95.90%, 91.90% and 0.8505.SVM diagnostic model after PCA-SMOTE oversampling technique, it is diagnostic
It can all be promoted.
Training accuracy is increased as screw-type water chiller sample set over-sampling ratio increases, when screw cooling-water machine
When group sample set over-sampling reaches 300%, training accuracy reaches highest, is 94.84%, but when over-sampling 400%, training
Accuracy is declined slightly.Test accuracy reaching highest in over-sampling 100%, be 96.70%, over-sampling 200%,
When 300% and 400%, test accuracy is declined slightly, but is held at 96% or more, after SMOTE oversampling technique
The robustness of SVM diagnostic model is preferable.
G-mean value is obtaining larger improvement after SMOTE oversampling technique, and G-mean value mentions when over-sampling 100%
Maximum is risen, promotes 1.57% compared with SVM model, in over-sampling 100%, G-mean value declines to a great extent, as over-sampling promotion is arrived
When 200%, G-mean value is risen again, and is still better than SVM model, when over-sampling 400%, G-mean value again under
Drop；MCC value variation tendency and G-mean value are identical, and when over-sampling 100%, MCC value reaches most preferably, then as this sample of over-sampling
The increase of this number, hence it is evident that it reduces, in over-sampling 300%%, is promoted again, but better than SVM model.
To sum up, the indices situation of change of assay diagnostic model overall performance, obtains PCA-SMOTE100%-
SVM model is excellent diagnostics model, and generalization ability and robustness are stronger, is suitble to the fault diagnosis of water cooler, can be generalized to
In the water cooler of other homology class units and different type of machines, it can reach preferable diagnosis performance.
All kinds of performance of fault diagnosis
In Fig. 5,6,7, Normal is normal status data, and RefLeak is leakage of refrigerant, ReduCF condenser water flow
Deficiency, RefOver overcharge for refrigerant, and Oversampling ratio is over-sampling multiplying power, and Sensitivity is recall ratio,
Precision is precision ratio, and it is precision ratio (Precision) and recall ratio that F-measure value, which is a kind of statistic,
(Sensitivity) weighted harmonic mean.
As shown in Fig. 5,6,7, using recall ratio of the diagnostic model of PCA-SMOTE oversampling technique in all kinds of failures,
Precision ratio and F-measure value are all relatively better than original SVM model.
Wherein, Sensitivity-Normal most preferably PCA-SMOTE200%-SVM model is 95.30%, than SVM mould
Type increases by 2.8%；Five class model performance capabilities of Sensitivity-RefLeak and Sensitivity-ReduCF is identical；
Sensitivity-RefOver most preferably PCA-SMOTE100%-SVM model is worth and is promoted for 96.20% compared with archetype
1.3%.
PCA-SMOTE100%-SVM model is preferable in Precision-Normal and Precision-ReduCF performance, especially
It is Precision-ReduCF, promotes 2.34% compared with SVM model；It is PCA- that Precision-RefOver, which puts up the best performance,
SMOTE200%-SVM model is promoted to 99.47% from the 98.75% of SVM model.Five kinds of model performance phases of RefLeak failure
Together.
For F-measure value, F-measure-ReduCF and F-measure-RefOver are respectively from SVM model
93.02%, 96.79% it is promoted to the 94.33% and 97.52% of PCA-SMOTE100%-SVM, is five kinds of model performance capabilities
Best model；And it is PCA-SMOTE400%-SVM model that F-measure-Normal, which puts up the best performance, is 97.29%, compared with SVM
Model promotes 1.34%.
To sum up, in five kinds of models, PCA-SMOTE100%-SVM is preferable, equal to Normal, ReduCF and RefOver failure
There is promotion.Still have in the case that in the case where unbalance factor is up to 10%, screw-type water chiller sample size is minimum compared with
Good diagnosis performance, it is seen that the method for diagnosing faults based on PCA-SMOTE-SVM is practicable.Since training set accounts for main part
The data set divided is centrifugal refrigerating machines, and fraction data set is screw-type water chiller, and PCA-SMOTE-SVM model remains unchanged
Higher diagnosis performance is kept, can prove that the diagnostic model can be used in different types of water cooler, diagnostic model pushes away
Wide property is improved.
The action and effect of embodiment
The water cooler method for diagnosing faults according to involved in the present embodiment is mentioned using synthesis minority class oversampling technique
A kind of water cooler method for diagnosing faults based on PCA-SMOTE-SVM algorithm has been supplied, it can be to different types of water cooler
Fault diagnosis is carried out, is had broad application prospects in the practical application of water cooler fault diagnosis technology.
Above embodiment is preferred case of the invention, the protection scope being not intended to limit the invention.
Claims (5)
1. a kind of water cooler method for diagnosing faults, which comprises the following steps:
S1 carries out fault simulation experiment to screw-type water chiller and acquires screw-type water chiller fault simulation experimental data,
The screw-type water chiller fault simulation experimental data includes normal data and fault simulation data,
Fault simulation experiment is carried out to centrifugal refrigerating machines, is formed after acquiring centrifugal refrigerating machines fault simulation experimental data
First training set data, the centrifugal refrigerating machines fault simulation experimental data include normal data and fault simulation data；
S2, it is random after steady state process and removal noise processed to the screw-type water chiller fault simulation experimental data
Selection the second training set data of composition and test set data；
S3 establishes original training set data, which includes first training set data and second instruction
Practice collection data；
S4 after carrying out over-sampling processing to the original training set data, obtains the over-sampling training sample of different over-sampling multiplying powers
This collection data；
S5 is trained the over-sampling training sample set data using SVM model, obtains diagnostic model；
The test set is put into the diagnostic model and is tested by S6, obtains the diagnosis knot of different water cooler fault models
Fruit.
2. water cooler method for diagnosing faults according to claim 1, which is characterized in that further comprising the steps of:
S7 assesses the diagnostic model using different evaluation methods, obtains optimal water cooler imbalance sampling
When excellent diagnostics model.
3. water cooler method for diagnosing faults according to claim 1 or 2, it is characterised in that:
Wherein, the screw cooling-water machine normal data include group an evaporator leaving water temperature data, condenser inflow temperature data,
Spool position data and frequency data,
The screw-type water chiller fault simulation data include refrigerant leak data, condenser water flow deficiency data and system
Cryogen overcharges data.
4. water cooler method for diagnosing faults according to claim 1, it is characterised in that:
Wherein, first training set data include normal condition, leakage of refrigerant, condenser fouling, condenser water flow not
Foot, condenser containing incoagulable gas, evaporator water flow is insufficient, refrigerant overcharges and excessive each 1000 sample numbers of lubricating oil
According to.
5. water cooler method for diagnosing faults according to claim 1, it is characterised in that:
Wherein, the expression formula of the SVM model are as follows:
In formula, sgn () is sign function, K (x, xi) it is to select RBF function, i.e.,
In formula: σ is the width of radial basis function, and σ is smaller, and the width of radial basis function is smaller, more selective,It is
Radial base nuclear parameter, g is bigger, and radial basis function is more selective, and x is to differentiate sample, xiFor training sample, i is number of samples,For the solution of dual problem, b*For threshold value, N is sample total.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811516128.8A CN109540562A (en) | 2018-12-12 | 2018-12-12 | A kind of water cooler method for diagnosing faults |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811516128.8A CN109540562A (en) | 2018-12-12 | 2018-12-12 | A kind of water cooler method for diagnosing faults |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109540562A true CN109540562A (en) | 2019-03-29 |
Family
ID=65854393
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811516128.8A Pending CN109540562A (en) | 2018-12-12 | 2018-12-12 | A kind of water cooler method for diagnosing faults |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109540562A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110333093A (en) * | 2019-07-19 | 2019-10-15 | 东莞市泰丰空调制冷设备有限公司 | A kind of water cooler test platform |
| CN110388773A (en) * | 2019-07-15 | 2019-10-29 | 重庆美的通用制冷设备有限公司 | Fault detection method, system and the water cooler of water cooler |
| CN110702438A (en) * | 2019-09-23 | 2020-01-17 | 天津大学 | Centrifugal water chilling unit and fresh air system combined fault diagnosis method |
| CN110986407A (en) * | 2019-11-08 | 2020-04-10 | 杭州电子科技大学 | Fault diagnosis method for centrifugal water chilling unit |
| CN112364999A (en) * | 2020-10-19 | 2021-02-12 | 深圳市超算科技开发有限公司 | Training method and device for water chiller adjustment model and electronic equipment |
| CN113283532A (en) * | 2021-06-09 | 2021-08-20 | 重庆大学 | Water chilling unit fault diagnosis method based on deep migration learning |
| CN115127194A (en) * | 2022-08-25 | 2022-09-30 | 蘑菇物联技术（深圳）有限公司 | Method, equipment and medium for detecting faults of central air-conditioning water chilling unit |
| CN116499748A (en) * | 2023-06-27 | 2023-07-28 | 昆明理工大学 | Bearing fault diagnosis method and system based on improved SMOTE and classifier |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480715A (en) * | 2017-08-10 | 2017-12-15 | 合肥工业大学 | The method for building up and system of the transmission device failure predication model of hydroforming equipment |
| CN107563435A (en) * | 2017-08-30 | 2018-01-09 | 哈尔滨工业大学深圳研究生院 | Higher-dimension unbalanced data sorting technique based on SVM |
| CN107832875A (en) * | 2017-10-27 | 2018-03-23 | 云南电网有限责任公司电力科学研究院 | A kind of transmission line lightning stroke failure prediction method of improved adaptive GA-IAGA Support Vector Machines Optimized |
| CN108376264A (en) * | 2018-02-26 | 2018-08-07 | 上海理工大学 | A kind of handpiece Water Chilling Units method for diagnosing faults based on support vector machines incremental learning |
| US20180322958A1 (en) * | 2017-05-04 | 2018-11-08 | Efthymios Kalafatis | Machine learning, natural language processing and network analysis-guided discovery related to medical research |
-
2018
- 2018-12-12 CN CN201811516128.8A patent/CN109540562A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180322958A1 (en) * | 2017-05-04 | 2018-11-08 | Efthymios Kalafatis | Machine learning, natural language processing and network analysis-guided discovery related to medical research |
| CN107480715A (en) * | 2017-08-10 | 2017-12-15 | 合肥工业大学 | The method for building up and system of the transmission device failure predication model of hydroforming equipment |
| CN107563435A (en) * | 2017-08-30 | 2018-01-09 | 哈尔滨工业大学深圳研究生院 | Higher-dimension unbalanced data sorting technique based on SVM |
| CN107832875A (en) * | 2017-10-27 | 2018-03-23 | 云南电网有限责任公司电力科学研究院 | A kind of transmission line lightning stroke failure prediction method of improved adaptive GA-IAGA Support Vector Machines Optimized |
| CN108376264A (en) * | 2018-02-26 | 2018-08-07 | 上海理工大学 | A kind of handpiece Water Chilling Units method for diagnosing faults based on support vector machines incremental learning |
Non-Patent Citations (2)
| Title |
|---|
| 胡云鹏: "基于主元分析的冷水机组传感器故障检测效率研究", 《CNKI中国博士学位论文全文数据库工程科技Ⅱ辑》 * |
| 韩华: "基于顺序集成方法的制冷系统故障检测与诊断研究", 《CNKI中国博士学位论文全文数据库工程科技Ⅱ辑》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110388773A (en) * | 2019-07-15 | 2019-10-29 | 重庆美的通用制冷设备有限公司 | Fault detection method, system and the water cooler of water cooler |
| WO2021008176A1 (en) * | 2019-07-15 | 2021-01-21 | 重庆美的通用制冷设备有限公司 | Fault detection method and system for water chilling unit, and water chilling unit |
| CN110333093A (en) * | 2019-07-19 | 2019-10-15 | 东莞市泰丰空调制冷设备有限公司 | A kind of water cooler test platform |
| CN110702438A (en) * | 2019-09-23 | 2020-01-17 | 天津大学 | Centrifugal water chilling unit and fresh air system combined fault diagnosis method |
| CN110702438B (en) * | 2019-09-23 | 2020-09-18 | 天津大学 | Centrifugal water chilling unit and fresh air system combined fault diagnosis method |
| CN110986407A (en) * | 2019-11-08 | 2020-04-10 | 杭州电子科技大学 | Fault diagnosis method for centrifugal water chilling unit |
| CN112364999A (en) * | 2020-10-19 | 2021-02-12 | 深圳市超算科技开发有限公司 | Training method and device for water chiller adjustment model and electronic equipment |
| CN113283532A (en) * | 2021-06-09 | 2021-08-20 | 重庆大学 | Water chilling unit fault diagnosis method based on deep migration learning |
| CN113283532B (en) * | 2021-06-09 | 2022-09-09 | 重庆大学 | Water chilling unit fault diagnosis method based on deep migration learning |
| CN115127194A (en) * | 2022-08-25 | 2022-09-30 | 蘑菇物联技术（深圳）有限公司 | Method, equipment and medium for detecting faults of central air-conditioning water chilling unit |
| CN115127194B (en) * | 2022-08-25 | 2022-11-25 | 蘑菇物联技术（深圳）有限公司 | Method, equipment and medium for detecting faults of central air conditioner water chilling unit |
| CN116499748A (en) * | 2023-06-27 | 2023-07-28 | 昆明理工大学 | Bearing fault diagnosis method and system based on improved SMOTE and classifier |
| CN116499748B (en) * | 2023-06-27 | 2023-08-29 | 昆明理工大学 | Bearing fault diagnosis method and system based on improved SMOTE and classifier |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109540562A (en) | A kind of water cooler method for diagnosing faults | |
| Fan et al. | Chiller fault diagnosis with field sensors using the technology of imbalanced data | |
| CN105300692B (en) | A kind of bearing failure diagnosis and Forecasting Methodology based on expanded Kalman filtration algorithm | |
| CN102955902B (en) | Method and system for evaluating reliability of radar simulation equipment | |
| Shen et al. | Augmented data driven self-attention deep learning method for imbalanced fault diagnosis of the HVAC chiller | |
| CN107942854A (en) | A kind of automatic clinical chemistry analyzer remote device diagnostics strategy | |
| CN110674120A (en) | Wind power plant data cleaning method and device | |
| CN114386537B (en) | Lithium battery fault diagnosis method and device based on Catboost and electronic equipment | |
| CN108758969A (en) | A kind of handpiece Water Chilling Units fault detection method and system | |
| CN115638875B (en) | Power plant equipment fault diagnosis method and system based on map analysis | |
| CN111639850A (en) | Quality evaluation method and system for multi-source heterogeneous data | |
| CN111311135A (en) | Transformer substation energy efficiency assessment method | |
| CN108491503A (en) | A kind of method and system for the fault type determining transformer based on data analysis | |
| CN113469252A (en) | Extra-high voltage converter valve operation state evaluation method considering unbalanced samples | |
| CN112380763A (en) | System and method for analyzing reliability of in-pile component based on data mining | |
| CN103136440A (en) | Method and device of data processing | |
| CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
| CN116319255A (en) | Root cause positioning method, device, equipment and storage medium based on KPI | |
| CN110110784A (en) | A kind of transformer fault discrimination method based on transformer correlation operation data | |
| CN109116833A (en) | Based on improvement drosophila-bat algorithm mechanical failure diagnostic method | |
| CN114722641A (en) | Lubricating oil state information integrated evaluation method and system for detection laboratory | |
| CN113204894A (en) | Construction method and application of electric energy metering abnormity diagnosis model | |
| CN110874088A (en) | Monitoring method of ship key equipment system based on multi-dimensional vector model | |
| Xie et al. | Knowledge Acquisition for Transformer Condition Assessment Using Synthetic Minority Over-sampling Technique and Decision Tree Algorithm | |
| Yu et al. | Center loss guided prototypical networks for unbalance few-shot industrial fault diagnosis |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190329 |