CN117404765B - Air conditioner system fan fault diagnosis method and system under weak supervision condition and air conditioner - Google Patents
Air conditioner system fan fault diagnosis method and system under weak supervision condition and air conditioner Download PDFInfo
- Publication number
- CN117404765B CN117404765B CN202311714348.2A CN202311714348A CN117404765B CN 117404765 B CN117404765 B CN 117404765B CN 202311714348 A CN202311714348 A CN 202311714348A CN 117404765 B CN117404765 B CN 117404765B
- Authority
- CN
- China
- Prior art keywords
- data
- task
- signal
- fan
- layer
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000003745 diagnosis Methods 0.000 title claims abstract description 25
- 238000004378 air conditioning Methods 0.000 claims abstract description 43
- 238000012360 testing method Methods 0.000 claims description 38
- 238000012549 training Methods 0.000 claims description 37
- 238000012545 processing Methods 0.000 claims description 25
- 230000004927 fusion Effects 0.000 claims description 20
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 19
- 238000011156 evaluation Methods 0.000 claims description 18
- 238000001514 detection method Methods 0.000 claims description 17
- 238000010606 normalization Methods 0.000 claims description 16
- 238000005516 engineering process Methods 0.000 claims description 15
- 238000005070 sampling Methods 0.000 claims description 13
- 238000009826 distribution Methods 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 238000007500 overflow downdraw method Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 239000013589 supplement Substances 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 2
- 238000012423 maintenance Methods 0.000 abstract description 3
- 101100272279 Beauveria bassiana Beas gene Proteins 0.000 description 6
- 238000010586 diagram Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 101100421299 Schizosaccharomyces pombe (strain 972 / ATCC 24843) set7 gene Proteins 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/30—Control or safety arrangements for purposes related to the operation of the system, e.g. for safety or monitoring
- F24F11/32—Responding to malfunctions or emergencies
- F24F11/38—Failure diagnosis
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
- F24F11/63—Electronic processing
- F24F11/64—Electronic processing using pre-stored data
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0895—Weakly supervised learning, e.g. semi-supervised or self-supervised learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Mechanical Engineering (AREA)
- Combustion & Propulsion (AREA)
- Signal Processing (AREA)
- Chemical & Material Sciences (AREA)
- Fuzzy Systems (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to the field of intelligent household equipment fault diagnosis, in particular to a method, a system and an air conditioner for diagnosing an air conditioner system fan fault under a weak supervision condition. The invention solves the problem of difficult automatic diagnosis of the fan faults of the air conditioning system under different working conditions by utilizing small sample learning, and brings great help to the operation of maintenance personnel.
Description
Technical Field
The invention relates to the field of intelligent household fault diagnosis, in particular to a method and a system for diagnosing fan faults of an air conditioning system under weak supervision conditions and an air conditioner.
Background
In the traditional air conditioning system, the performance of the automatic detection method for the fan faults of the air conditioning system is difficult to improve because the number of sensors is small and the quality is low. With the rapid development of technologies such as artificial intelligence, cloud computing, sensors and the like, more and more intelligent home systems are entering our lives, in particular air conditioning systems. An air conditioning system is a complex operation system, various problems often occur in long-term operation, and particularly, each component of a fan is prone to failure. However, the air conditioning system has a complex structure, and the elimination of fan faults can consume a large amount of manpower and material resources by simply relying on manual field detection, so that the problem of low efficiency exists, and the professional requirement on maintenance personnel is higher.
The highest accuracy rate in the existing automatic diagnosis method for fan faults of the air conditioning system is based on a data driving method, but the method needs a large amount of training samples and cannot realize generalization on different data distribution. Therefore, aiming at the problems of complex operation condition of an air conditioning system, various fan faults and difficult overall fault data acquisition, the invention discloses a method and a system for diagnosing the fan faults of the air conditioning system under the weak supervision condition and a technical scheme of an air conditioner.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a method and a system for diagnosing the fan faults of an air conditioning system under the weak supervision condition and an air conditioner, solves the problem of difficult automatic diagnosis of the fan faults of the air conditioning system under different working conditions, provides a new technical scheme for automatic diagnosis of the fan faults of the air conditioning system, and brings great help to the operation of maintenance personnel.
The technical scheme for solving the technical problems is as follows: a fan fault diagnosis method for an air conditioning system under weak supervision condition comprises the following steps:
s1, continuously acquiring operation parameters of a fan of an air conditioning system under a working condition A and a working condition B respectively, wherein the operation parameters comprise a state signal and a temperature signal of the fan;
s2, preprocessing the data acquired in the step S1, processing the state signals and the temperature signals of each fan according to a maximum and minimum normalization technology, calculating the number N of sample points contained in each state signal by acquiring the state signals of each sensor of the fan of the air conditioning system, and performing data splicing processing on the samples of the obtained state signals and the temperature signals;
s3, carrying out data expansion processing on the data preprocessed in the step S2, carrying out window division on samples spliced by the data, generating a split point signal on the obtained signal, and then carrying out sample data fusion and dimension conversion;
s4, respectively constructing N-way K-shot task level data sets under the working condition A and the working condition B, wherein the data sets are as followsPersonal category and->N-way K-shot task data are divided from samples, namely +.>Randomly selecting N-way categories as target categories, constructing a support set and a query set under a single task to generate a data set, repeatedly generating the data set to form a task-level small sample data set, and then dividing the data set;
s5, constructing a multi-level small sample data feature extractor by extracting level features, multi-level multi-granularity feature fusion and an LSTM layer and a full connection layer according to the data set in the S4;
and S6, performing super-parameter setting on the feature extractor constructed in the step S5, performing evaluation and generalization on the parameter fine adjustment of the feature extractor by utilizing the training data of the working condition B, and performing test set evaluation on the training data of the working condition A through the multi-level feature extractor.
Further, S1 specifically includes the following procedures:
s1.1, collecting fan state signals: the vibration sensor is used for collecting all state signals of the fan under all working conditions, and the fan comprises the following state signals: normal stateMain shaft failure->Bearing inner race failure->Bearing outer race failure->Bearing Rolling element failure->Blade failure->Gear case failure->The fan status signals under working conditions A and B are marked as +.>And->Wherein->，；
S1.2, acquiring fan temperature signals: temperature sensors are utilized to collect temperatures of fans of air conditioning system under all working conditionsThe sampling rate is 60Hz, and the temperature signals under the working conditions A and B are marked as +.>And->。
Further, S2 specifically includes the following procedures: as described in step S1, each sensor of the fan of the air conditioning system continuously operates for M minutes to collect fan status signals, each status signal including n=in totalSample points;
s2.1 maximum minimum normalization: processing the state signals and the temperature signals of each fan according to a maximum and minimum normalization technology, wherein the calculation formula is as follows:
，
wherein,represents a status signal or temperature signal, min (++>) Representation->Max (++>) Representation->Maximum value of>Representing the normalized signal;
s2.2, data segmentation and sample division: the N sample points contained in each state signal are segmented according to the length, the length is recorded as lens, each state contains m samples,；
s2.3, data splicing: the temperature signal and the processed state signal are subjected to data splicing,
，
wherein,a signal representing the fusion of the status signal and the temperature signal, and (2)>Representing the normalized state signal, +.>A signal representing the normalized temperature signal.
Further, the specific steps of S3 are as follows:
3.1 Dividing window signals: from the slaveThe first sampling point of the signal starts to divide the window according to the window w and the step length s, and the number of the divided windows is +.>=n-w，/>The signal after the signal dividing window is
；
3.2 Generating a component site signal: calculating window signalsCorresponding first and third subsites, obtaining the first subsite signal +.>And a sliding window third quantile signal +.>By->And->The w/2 0 values of each of the front and rear supplements are kept +.>、/>And->Is consistent in dimension;
3.3 Sample data fusion and dimension conversion:
will be、/>And->Fusion is performed to generate new sample data +.>I.e.
，
Wherein,is +.>，/>Representing a feature fusion operation;
the 1-dimensional signal is then applied using Reshape technologyBecome 2-dimensional signal->，/>Is +.>。
Further, S4 specifically includes the following procedures:
s4.1, generating a support set and a query set of single tasks: from the slaveN-way categories are randomly selected as target categories, and a support set and a query set sample under the task are constructed, wherein the support set and the query set sample are specifically as follows:
(1) For the ith category in N-way, randomly selecting K-shot+K-qry samples from m samples corresponding to the category, wherein the former K-shot samples are support setsK-qry samples later are query set +.>N-way represents the number of samples, K-shot represents the number of samples of the support set, K-qry represents the number of samples of the query set, i represents the number of categories, and m represents the number of samples;
(2) Repeating the steps, iterating the N-way times until the support sets of N-way target categories are selectedAnd query set->；
(3) Generating support sets under a single taskAnd query set->The method comprises the following steps of:
，
;
(4) MergingAnd->Generating a data set under a single task +.>；
S4.2 generates a task-level small sample dataset: repeatedly executing S4.1 until enough N-way and K-shot small sample task data sets are generated, and combining the generated T data setsTo form a task-level small sample data set, T represents the number of data sets under a single task, and the task-level small sample data set D is: />；
S4.3, dividing training sets and test sets: dividing D into training sets according to proportionAnd test set->。
Further, S5 specifically includes the following procedures: building a residual network model based on 3 'convolution blocks and jump connection', wherein each convolution block comprisesContaining，/>And->Is a convolution layer of (1) bulk processing normalization layer->And a nonlinear layer->。
S5.1, extracting hierarchical features:
(1) Will beAs input to the first residual block, first through the 3 stacked convolutional layers of the first convolutional block, one +.>Layer and->A layer, a maximum pooling layer obtaining the output of the first layer>Then will->The 3 stacked convolutional layers, one +.>Layer and->After the layer the output of the second convolution block is obtained>Finally, using the jump connection will +.>And->Adding to obtain the output of the first residual block>；
(2) Will beAs input to the second residual block, first through the 3 stacked convolutional layers of the first convolutional block, one +.>Layer and->The layer gets the output of the first convolution block +.>Then will->The 3 stacked convolutional layers, one +.>Layer and->After the layer the output of the second convolution block is obtained>Finally, using the jump connection will +.>And->Adding to obtain the output of the second residual block +.>；
(3) Will beAs input to the third residual block, first through the 3 stacked convolutional layers of the first convolutional block, one +.>Layer and->The layer gets the output of the first convolution block +.>Then will->The 3 stacked convolutional layers, one +.>Layer and->After the layer the output of the second convolution block is obtained>Finally, using the jump connection will +.>And->Adding to obtain the output of the third residual block +.>；
S5.2, multi-layer multi-granularity characteristic fusion: will beSplicing to extract multi-level and multi-granularity feature map to form multi-level feature vector +.>：
；
S5.3 LSTM and full connectivity layer: by combining the feature vectors obtained in S5.2Inputting to 2 cascaded LSTM layers to extract features with long dependency +.>And utilizes two fully connected layers +.>The extraction of global features is realized, and the operation features of the air conditioner fan under the working condition A are obtained>The method comprises the following steps: />，
。
Further, S6 specifically includes the following procedures:
s6.1 training parameter setting: performing super-parameter setting on the feature extractor constructed in the step S5;
s6.2 model training:
(1) Will beThe features are extracted by inputting the features into the multi-level feature extractor constructed in the step S5, and the support set under a certain single task is +.>And query set->Respectively outputting corresponding support set multi-level characteristics +.>And query set multi-level feature->;
(2) By using Euclidean distanceTo measure the similarity and calculate the similarity of the two>The method comprises the following steps:
(3) The calculated Sim and the real labelInput to negative log likelihood loss function->The loss under the current single task is calculated>The method comprises the following steps:
loss valueComparing with the initial optimum loss value if the loss value +.>If the initial optimal loss value is reduced, updating parameters of the feature extractor, and saving the model, otherwise, not updating;
(4) Repeating the above process until a predetermined end condition is reached;
s6.3 model evaluation and generalization:
(1) Will beInputting the test set into the stored model to obtain multi-level features corresponding to the test setAnd->Then calculating the similarity Sim of the two by using the Euclidean distance, and finally distributing the label corresponding to the position with the maximum similarity to the test set data to realize the automatic detection of the fan faults and obtain the test set evaluation result under the working condition A;
(2) And (3) sequentially executing steps S2, S3 and S4 on the data acquired under the working condition B, dividing a training set and a test set, performing fine adjustment on the parameters of the feature extractor stored in the step S6.2 by using the training set, and then evaluating on the test set to realize generalization of model evaluation among different data distribution.
The invention also provides a fan fault diagnosis system of the air conditioning system under the weak supervision condition, which comprises the following modules:
and a data acquisition module: collecting operation data of an air conditioner system fan under a working condition A and a working condition B;
and a data processing module: preprocessing the data acquired by the data acquisition module, carrying out maximum and minimum normalization, segmentation and splicing on the data, and then transmitting the processed data to the data expansion module;
and a data expansion module: the data dimension is expanded by utilizing a sliding window partitioning point technology, the diversity of fault data is increased, and a multi-level and multi-granularity feature extractor is constructed by utilizing a level feature fusion method;
a data set processing module: after a small sample data set is constructed by using a meta task divider, training is carried out by using a support set sample and a query set sample, and optimized parameter information of the feature extractor is obtained;
test set processing module: and (3) carrying out test set evaluation on the training data of the working condition A through a multi-level feature extractor, and carrying out fine adjustment on parameters of the feature extractor by utilizing the training data of the working condition B, so as to evaluate the training data on the test set to achieve generalization of the fault automatic diagnosis method among different data distributions.
The invention also provides an air conditioner, which comprises the air conditioner system fan fault diagnosis system under the weak supervision condition, and the air conditioner running executes the air conditioner system fan fault diagnosis method under the weak supervision condition.
The technical scheme has the following advantages or beneficial effects:
the invention provides a fan fault diagnosis, a system and an air conditioner of an air conditioning system under a weak supervision condition, and aims to realize automatic fault detection of the fan of the air conditioning system and generalization of a diagnosis model under multiple working conditions.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.
The invention is to be interpreted as illustrative and not in a limiting sense.
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a block flow diagram of a convolution of the present invention.
Fig. 3 is a diagram of a multi-level feature extraction architecture of the present invention.
Fig. 4 shows the signal-to-noise ratio under condition a.
Fig. 5 is a signal-to-noise ratio under condition B.
Detailed Description
In order to clearly illustrate the technical features of the present solution, the present invention will be described in detail below with reference to the following detailed description and the accompanying drawings. The following disclosure provides many different embodiments, or examples, for implementing different structures of the invention. In order to simplify the present disclosure, components and arrangements of specific examples are described below. Furthermore, the present invention may repeat reference numerals and/or letters in the various examples. This repetition is for the purpose of simplicity and clarity and does not in itself dictate a relationship between the various embodiments and/or configurations discussed. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
Embodiment 1 as shown in fig. 1, a fan fault diagnosis method for an air conditioning system under a weak supervision condition includes the following steps:
s1, continuously acquiring operation parameters of a fan of an air conditioning system under a working condition A and a working condition B respectively, wherein the operation parameters comprise a state signal and a temperature signal of the fan;
s2, preprocessing the data acquired in the step S1, processing the state signals and the temperature signals of each fan according to a maximum and minimum normalization technology, calculating the number N of sample points contained in each state signal by acquiring the state signals of each sensor of the fan of the air conditioning system, and performing data splicing processing on the samples of the obtained state signals and the temperature signals;
s3, carrying out data expansion processing on the data preprocessed in the step S2, carrying out window division on samples spliced by the data, generating a split point signal on the obtained signal, and then carrying out sample data fusion and dimension conversion;
s4, respectively constructing N-way K-shot task level data sets under the working condition A and the working condition B, wherein the data sets are as followsPersonal category and->N-way K-shot task data are divided from samples, namely +.>Randomly selecting N-way categories as target categories, constructing a support set and a query set under a single task to generate a data set, repeatedly generating the data set to form a task-level small sample data set, and then dividing the data set;
s5, constructing a multi-level small sample data feature extractor by extracting level features, multi-level multi-granularity feature fusion and an LSTM layer and a full connection layer according to the data set in the S4;
and S6, performing super-parameter setting on the feature extractor constructed in the step S5, performing evaluation and generalization on the parameter fine adjustment of the feature extractor by utilizing the training data of the working condition B, and performing test set evaluation on the training data of the working condition A through the multi-level feature extractor.
Further, S1 specifically includes the following procedures:
s1.1, collecting fan state signals: the vibration sensor is used for collecting all state signals of the fan under all working conditions, and the fan comprises the following state signals: normal stateMain shaft failure->Bearing inner race failure->Bearing outer race failure->Bearing Rolling element failure->Blade failure->Gear case failure->The fan status signals under working conditions A and B are marked as +.>And->The sampling rate is 12kHz, wherein +.>，/>；
S1.2, acquiring fan temperature signals: temperature sensors are utilized to collect temperatures of fans of air conditioning system under all working conditionsThe sampling rate is 60Hz, and the temperature signals under the working conditions A and B are marked as +.>And->。
Further, S2 specifically includes the following procedures: as described in step S1, each sensor of the fan of the air conditioning system continuously operates for M minutes to collect fan status signals, each status signal including n=in totalSetting M=1 to obtain the number of sample points N= 720000 contained in each state signal;
s2.1 maximum minimum normalization: processing the state signals and the temperature signals of each fan according to a maximum and minimum normalization technology, wherein the calculation formula is as follows:
wherein,represents a status signal or temperature signal, min (++>) Representation->Max (++>) Representation->Maximum value of>Representing the normalized signal;
s2.2, data segmentation and sample division: the N sample points contained in each state signal are segmented according to the length, the length is recorded as lens, each state contains m samples,；
as is clear from S1.2, n= 720000, and lens=2000 is set to obtain the number of samples m=360; co-inclusion in a segmented data set7 states, each with 360 samples, total->Samples.
S2.3, data splicing: the influence on the classification performance during signal acquisition or division is removed by removing 37 sampling points before and after 2000 sampling points of the division to obtain a sample with the length of 1926, then the temperature signal and the processed state signal are subjected to data splicing,
wherein,representing status signals and temperature signalsFused signal,/->A second signal representing the normalized state signal, < >>Representing the normalized temperature signal, obtaining a signal of 1/6 second as one sample according to the sampling rate of the temperature signal being 60Hz, the sampling rate of the state signal being 10kHz and the length lens being 2000, knowing that 60 sampling points can be acquired in one second by the sampling rate of the temperature signal being 60Hz, determining that 10 sampling points are acquired in 1/6 second, and finally obtaining%>The signal contains 1926+10=1936 sample points.
Further, the specific steps of S3 are as follows:
3.1 Dividing window signals: from the slaveThe first sampling point of the signal starts to divide the window according to the window w and the step length s, and the number of the divided windows is +.>N-w, n is 1936, w=10, s=1 is set, and +.>=1926，/>The signal after the signal dividing window is +.>；
3.2 Generating a component site signal: calculating window signalsCorresponding first and third subsites, obtaining the first subsite signal +.>And a sliding window third quantile signal +.>，/>And->After window division, the window is more than original +.>The signal length is less than w points, in order to keep the lengths of the 3 signals uniform, by the method of +.>Andthe w/2 0 values of each of the front and rear supplements are kept +.>、/>And->Is consistent in dimension;
3.3 Sample data fusion and dimension conversion:
will be、/>And->Fusion is performed to generate new sample data +.>I.e.
Wherein,is +.>，/>Representing a feature fusion operation;
the 1-dimensional signal is then applied using Reshape technologyBecome 2-dimensional signal->，/>Is +.>。
Further, S4 specifically includes the following procedures:
s4.1, generating a support set and a query set of single tasks: from the slaveN-way categories are randomly selected as target categories, and a support set and a query set sample under the task are constructed, wherein the support set and the query set sample are specifically as follows:
(1) For the ith category in N-way, randomly selecting K-shot+K-qry samples from m samples corresponding to the category, wherein the former K-shot samples are support setsK-qry samples later are query set +.>N-way represents the number of samples, K-shot represents the support setThe sample number, K-qry represents the sample number of the query set, N-way K-shot is a small sample learning and classifying method, i represents the category number, m represents the sample number, and N-way= 5,K-shot= 5,K-qry=15 is set;
(2) Repeating the steps, iterating the N-way times until the support sets of N-way target categories are selectedAnd query set->；
(3) Generating support sets under a single taskAnd query set->The method comprises the following steps of:
；
(4) MergingAnd->Generating a data set under a single task +.>；
S4.2 generates a task-level small sample dataset: repeatedly executing S4.1 until enough N-way and K-shot small sample task data sets are generated, and combining the generated T data setsTo construct a task-level small sample dataset, T representing the dataset under a single taskThe number, task-level small sample dataset D is:
；
s4.3, dividing training sets and test sets: dividing D into training sets according to the proportion of 7:3And test set->。
Further, as shown in fig. 2-3, S5 specifically includes the following processes: building a residual network model based on 3 'convolution blocks and jump connection', wherein each convolution block comprises，/>And->Is a convolution layer of (1) bulk processing normalization layer->And a nonlinear layer->。
S5.1, extracting hierarchical features:
(1) Will beAs input to the first residual block, first through the 3 stacked convolutional layers of the first convolutional block, one +.>Layer and->A layer, a maximum pooling layer to obtain a first layerOutput->Then will->The 3 stacked convolutional layers, one +.>Layer and->After the layer the output of the second convolution block is obtained>Finally, using the jump connection will +.>And->Adding to obtain the output of the first residual block>；
(2) Will beAs input to the second residual block, first through the 3 stacked convolutional layers of the first convolutional block, one +.>Layer and->The layer gets the output of the first convolution block +.>Then will->The 3 stacked convolutional layers, one +.>Layer and->After the layer the output of the second convolution block is obtained>Finally, using the jump connection will +.>And->Adding to obtain the output of the second residual block +.>；
(3) Will beAs input to the third residual block, first through the 3 stacked convolutional layers of the first convolutional block, one +.>Layer and->The layer gets the output of the first convolution block +.>Then will->The 3 stacked convolutional layers, one +.>Layer and->After the layer the output of the second convolution block is obtained>Finally, using the jump connection will +.>And->Adding to obtain the output of the third residual block +.>；
S5.2, multi-layer multi-granularity characteristic fusion: will beSplicing to extract multi-level and multi-granularity feature map to form multi-level feature vector +.>：
；
S5.3 LSTM and full connectivity layer: by combining the feature vectors obtained in S5.2Inputting to 2 cascaded LSTM layers to extract features with long dependency +.>And utilizes two fully connected layers +.>The extraction of global features is realized, and the operation features of the air conditioner fan under the working condition A are obtained>The method comprises the following steps:
，
，
the constructed multi-level feature extractor mainly comprises 3 residual network blocks, 2 cascaded LSTM layers and 2 fully connected layers, wherein each residual block contains KIs to process the normalization layer in batches>，Layer and->The convolution kernels K of the three residual layers are 32,64,128, respectively.
Further, S6 specifically includes the following procedures:
s6.1 training parameter setting: performing super-parameter setting on the feature extractor constructed in the step S5, wherein the optimizer selects an Adam optimizer, the learning rate is 0.01, and the loss function is a negative log likelihood loss functionBatch size 32;
s6.2 model training:
(1) Will beThe features are extracted by inputting the features into the multi-level feature extractor constructed in the step S5, and the support set under a certain single task is +.>And query set->Respectively outputting corresponding support set multi-level characteristics +.>And query set multi-level feature->;
(2) By using Euclidean distanceTo measure the similarity and calculate the similarity of the two>The method comprises the following steps:
，
(3) The calculated Sim and the real labelInput to negative log likelihood loss function->The loss under the current single task is calculated>The method comprises the following steps:
，
loss valueComparing with the initial optimum loss value if the loss value +.>If the initial optimal loss value is reduced, updating parameters of the feature extractor and saving the model, otherwise, not updating, wherein the initial optimal loss is set to 9999;
(4) Repeating the above process until reaching a predetermined end condition, setting the end condition to execute the above operation 1000 times;
s6.3 model evaluation and generalization:
(1) Will beInputting the test set into the stored model to obtain multi-level features corresponding to the test setAnd->Then calculating the similarity Sim of the two by using the Euclidean distance, and finally distributing the label corresponding to the position with the maximum similarity to the test set data to realize the automatic detection of the fan faults and obtain the test set evaluation result under the working condition A; FIG. 4 shows the results of automatic detection of fan faults under the conditions of signal to noise ratios of-4 db, -2db, 0db, 2db, 4db, 6db, 8db,10db and no noise, and the results in FIG. 4 show that the detection accuracy of the automatic detection technology provided by the invention can reach more than 97% under the condition of no noise. With the increase of the signal-to-noise ratio, the precision of the automatic detection technology tends to increase, and even when the signal-to-noise ratio is low, namely-4 db and-2 db, the precision of the automatic detection method reaches more than 86%, and the automatic detection method has stronger resistance to noise interference.
(2) Sequentially executing steps S2, S3 and S4 on the data acquired under the working condition B, dividing a training set and a test set, performing fine adjustment on parameters of the feature extractor stored in the step S6.2 by using the training set, and then evaluating on the test set to realize generalization of model evaluation among different data distributions; FIG. 5 shows the results of the fan failure automatic detection generalization under the working condition B, the signal to noise ratios are respectively-4, -2, 0, 2, 4,6, 8 and 10, and the fan failure automatic detection generalization under the noiseless condition, and as can be seen from the results in FIG. 5, the generalization performance of the automatic detection technology provided by the invention on a new data set can reach more than 94%, and the generalization precision of the automatic detection method provided by the invention can reach more than 85% even under the conditions of-4 db and-2 db with lower signal to noise ratio, so that the fan failure automatic detection with high precision can be realized by utilizing the fine adjustment of parameters under the condition that retraining data is not needed.
Embodiment 2 is an air conditioning system fan fault diagnosis system under weak supervision conditions, comprising the following modules:
and a data acquisition module: collecting operation data of an air conditioner system fan under a working condition A and a working condition B;
and a data processing module: preprocessing the data acquired by the data acquisition module, carrying out maximum and minimum normalization, segmentation and splicing on the data, and then transmitting the processed data to the data expansion module;
and a data expansion module: the data dimension is expanded by utilizing a sliding window partitioning point technology, the diversity of fault data is increased, and a multi-level and multi-granularity feature extractor is constructed by utilizing a level feature fusion method;
a data set processing module: after a small sample data set is constructed by using a meta task divider, training is carried out by using a support set sample and a query set sample, and optimized parameter information of the feature extractor is obtained;
test set processing module: and (3) carrying out test set evaluation on the training data of the working condition A through a multi-level feature extractor, and carrying out fine adjustment on parameters of the feature extractor by utilizing the training data of the working condition B, so as to evaluate the training data on the test set to achieve generalization of the fault automatic diagnosis method among different data distributions.
Embodiment 3 is an air conditioner comprising an air conditioning system fan fault diagnosis system under weak supervision, the air conditioner executing a method for air conditioning system fan fault diagnosis under weak supervision when operating.
According to the technical scheme, the method, the system and the air conditioner for diagnosing the fan faults of the air conditioning system under the weak supervision condition are provided, and the problem that the fan faults of the air conditioning system are difficult to obtain due to the fact that the operation working condition of the air conditioning system is complex and the fan faults are various and comprehensive fault data is solved.
While the foregoing description of the embodiments of the present invention has been presented with reference to the drawings, it is not intended to limit the scope of the invention, but rather, it is apparent that various modifications or variations can be made by those skilled in the art without the need for inventive work on the basis of the technical solutions of the present invention.
Claims (5)
1. A fan fault diagnosis method for an air conditioning system under weak supervision conditions is characterized by comprising the following steps:
s1, continuously acquiring operation parameters of a fan of an air conditioning system under a working condition A and a working condition B respectively, wherein the operation parameters comprise a state signal and a temperature signal of the fan;
s2, preprocessing the data acquired in the step S1, processing the state signals and the temperature signals of each fan according to a maximum and minimum normalization technology, calculating the number N of sample points contained in each state signal by acquiring the state signals of each sensor of the fan of the air conditioning system, and performing data splicing processing on the samples of the obtained state signals and the temperature signals;
s2 specifically comprises the following steps: according to the step S1, each sensor of the fan of the air conditioning system continuously operates for M minutes to collect fan status signals, and each status signal contains n=12000×m×60 sample points;
s2.1 maximum minimum normalization: processing the state signals and the temperature signals of each fan according to a maximum and minimum normalization technology, wherein the calculation formula is as follows:
wherein S is i Represents a status signal or a temperature signal, min (S i ) Represent S i Max (S) i ) Represent S i Maximum value of S' i Representing the normalized signal;
s2.2, data segmentation and sample division: carrying out segmentation processing on N sample points contained in each state signal according to the length, wherein the length is marked as kens, and each state contains m samples, and m=N/lens;
s2.3, data splicing: the temperature signal and the processed state signal are subjected to data splicing,
S sw ＝[S s ′；S w ′]，
wherein S is sw Representing the signal after the fusion of the state signal and the temperature signal, S s ' represents the normalized state signal, S w ' represents the normalized temperature signal;
s3, carrying out data expansion processing on the data preprocessed in the step S2, carrying out window division on samples spliced by the data, generating a split point signal on the obtained signal, and then carrying out sample data fusion and dimension conversion;
s3, the specific steps are as follows:
3.1 dividing window signal: from S sw The first sampling point of the signal starts to divide the window according to the window w and the step length s, and the number of the divided windows is N w N-w, n represents the nth sample after the fusion of the status signal and the temperature signal, S sw The signal after the signal dividing window is W sw ＝[s sw [0：0+w]；S sw [1：1+w]，...，S sw [N w ：N w +w]；
3.2 generating a quantile signal: calculate window signal W sw Corresponding first and third quantiles to obtain a sliding window first quantile signalAnd a sliding window third quantile signal +.>By at->And->The w/2 0 values of each of the front and rear supplements are kept +.>And S is sw Is consistent in dimension;
3.3 sample data fusion and dimension conversion:
will beAnd S is sw Fusing to generate new sample data S D I.e.
Wherein S is D Is of dimension R 3×1936 Concat represents a feature fusion operation;
the 1-dimensional signal S is then applied using the Reshape technique D Becomes a 2-dimensional signal S D ′，S D ' dimension R 3×44×44 ；
S4, respectively constructing N-way K-shot task level data sets under the working condition A and the working condition B, wherein the data sets are N C Category and N s N-way K-shot task data are divided from each sample, and N is used for dividing the data C Randomly selecting N-way categories as target categories, constructing a support set and a query set under a single task to generate a data set D task Repeatedly generating a data set to form a task-level small sample data set, and then dividing the data set;
s5, constructing a multi-level small sample data feature extractor by extracting level features, multi-level multi-granularity feature fusion, a long-period memory LSTM layer and a full-connection FC layer according to the data set in the S4;
s5 specifically comprises the following steps:
constructing a residual network model with 3 'convolution blocks+jump connection' as an infrastructure, wherein each convolution block comprises convolution layers of 1×1,3×3 and 1×1, and carrying out batch processing on a normalization layer BN and a nonlinear layer Relu;
s5.1, extracting hierarchical features:
(1) Will D task As input to the first residual block, the output F1 of the first layer is obtained first through the 3 stacked convolutional layers, one BN layer and one ReLu layer, one max-pooling layer of the first convolutional block cb1 The method comprises the steps of carrying out a first treatment on the surface of the Then F1 is added cb1 The 3 stacked convolution layers input to the second convolution block, one BN layer and one ReLu layer, to obtain the output F1 of the second convolution block cb2 The method comprises the steps of carrying out a first treatment on the surface of the Finally F1 is connected by jump cb1 And F1 to obtain the output F1 of the first residual block task ＝F1 cb1 +F1 cb2 ；
(2) Will F1 task As input to the second residual block, the output F2 of the first convolution block is first obtained by passing through the 3 stacked convolution layers of the first convolution block, one BN layer and one ReLu layer cb1 The method comprises the steps of carrying out a first treatment on the surface of the Then F2 is added cb1 The 3 stacked convolution layers input to the second convolution block, one BN layer and one ReLu layer, to obtain the output F2 of the second convolution block cb2 The method comprises the steps of carrying out a first treatment on the surface of the Finally F2 is connected by jump cb1 And F2 cb2 Adding to obtain the output F2 of the second residual block task ＝F2 cb1 +F2 cb2 ；
(3) F2 is to be F2 task As input to the third residual block, the output F3 of the first convolution block is first obtained by passing through the 3 stacked convolution layers of the first convolution block, one BN layer and one ReLu layer cb1 The method comprises the steps of carrying out a first treatment on the surface of the Then F3 is carried out cb1 The output F3 of the second convolution block is obtained after 3 stacked convolution layers, one BN layer and one ReLu layer are input into the second convolution block cb2 The method comprises the steps of carrying out a first treatment on the surface of the Finally F3 is connected by jump cb1 And F3 cb2 Adding to obtain the output F3 of the third residual block task ＝F3 cb1 +F3 cb2 ；
S5.2 multi-level multi-granularity feature fusion: will F1 task ，F2 task ，F3 task Stitching to extract multi-level and multi-granularity feature maps to form multi-level feature vectors
S5.3 LSTM and FC layer: by combining the feature vectors obtained in S5.2Inputting to 2 cascaded LSTM layers to extract features with long dependency +.>And the extraction of global features is realized by utilizing two full connection layers FC, so that the operation features of the air conditioner fan under the working condition A are obtained>Namely:
s6, performing super-parameter setting on the feature extractor constructed in the step S5, performing evaluation and generalization on the parameter fine adjustment of the feature extractor by using the training data of the working condition B, and performing test set evaluation on the training data of the working condition A by using the multi-level feature extractor;
s6 specifically comprises the following steps:
s6.1 training parameter setting: performing super-parameter setting on the feature extractor constructed in the step S5;
s6.2 model training:
(1) Will D tr Input to the multi-level feature extractor constructed in step S5 to extract features for a support set task under a single task support And query set task qry Respectively output corresponding support set multi-level characteristic FLL support And query set multi-level feature FLL qry ；
(2) Using Euclidean distance distance The similarity is measured, and the similarity Sim of the two is calculated as follows:
Sim＝Euclidean distance (FLL support ，FLL qry )，
(3) The Sim and the true label y obtained by calculation true The loss under the current single task is calculated by inputting the loss into the negative log likelihood loss function nllos as follows:
loss＝nlloss(Sim，y true )，
comparing the loss value loss with the initial optimal loss value, if the loss value loss is reduced compared with the initial optimal loss value, updating parameters of the feature extractor, and saving the model, otherwise, not updating;
(4) Repeating the above process until a predetermined end condition is reached;
s6.3 model evaluation and generalization:
(1) Will D test Inputting the test set into the stored model to obtain multi-level characteristic FLL corresponding to the test set support And FLL (flash light) qry Then calculating the similarity Sim of the two by using the Euclidean distance, and finally distributing the label corresponding to the position with the maximum similarity to the test set data to realize the automatic detection of the fan faults and obtain the test set evaluation result under the working condition A;
(2) And (3) sequentially executing steps S2, S3 and S4 on the data acquired under the working condition B, dividing a training set and a test set, performing fine adjustment on the parameters of the feature extractor stored in the step S6.2 by using the training set, and then evaluating on the test set to realize generalization of model evaluation among different data distribution.
2. The method for diagnosing fan failure of air conditioning system under weak supervision condition according to claim 1, wherein,
s1 specifically comprises the following steps:
s1.1, collecting fan state signals: collecting fans under various working conditions by using vibration sensorsEach status signal includes the following status signals: normal state S n Failure of main shaft S m Failure S of bearing inner race zi Failure S of outer ring of bearing zo Failure S of bearing rolling element zg Blade failure S 1 Failure S of gearbox c The fan status signals under the working conditions A and B are respectively marked as S sA And S is sB Wherein S is sA ，S sB ∈{S n ，S m ，S zi ，S zo ，S zg ，S l ，S c }；
S1.2, acquiring fan temperature signals: temperature S of each fan of air conditioning system under each working condition is acquired by using temperature sensor w The temperature signals under the working conditions A and B are respectively marked as S wA And S is wB 。
3. The method for diagnosing fan failure of air conditioning system under weak supervision condition according to claim 2, wherein,
s4 specifically comprises the following steps:
s4.1, generating a support set and a query set of single tasks: from N C N-way categories are randomly selected as target categories, and a support set and a query set sample under the task are constructed, wherein the support set and the query set sample are specifically as follows:
(1) For the ith category in N-way, randomly selecting K-shot+K-qry samples from m samples corresponding to the category, wherein the former K-shot samples are support setsK-qry samples after are query set +.>N-way represents the number of samples, K-shot represents the number of samples of the support set, K-qry represents the number of samples of the query set, i represents the number of categories, and m represents the number of samples;
(2) Repeating the steps, iterating the N-way times until the support sets of N-way target categories are selectedAnd query set->
(3) Generating a support set task under a single task support And query set task qry The method comprises the following steps of:
(4) Merging task support And task qry Generating a data set D under a single task task ；
S4.2 generates a task-level small sample dataset: repeatedly executing S4.1 until enough N-way and K-shot small sample task data sets are generated, and combining the generated T D task To form a task-level small sample data set, T represents the number of data sets under a single task, and the task-level small sample data set D is:
s4.3, dividing training sets and test sets: dividing D into training sets D according to proportion tr And test set D test 。
4. A fan fault diagnosis system for an air conditioning system under weak supervision condition, which performs the method for diagnosing fan fault of an air conditioning system under weak supervision condition as set forth in any one of claims 1 to 3, comprising the following modules:
and a data acquisition module: collecting operation data of an air conditioner system fan under a working condition A and a working condition B;
and a data processing module: preprocessing the data acquired by the data acquisition module, carrying out maximum and minimum normalization, segmentation and splicing on the data, and then transmitting the processed data to the data expansion module;
and a data expansion module: the data dimension is expanded by utilizing a sliding window partitioning point technology, the diversity of fault data is increased, and a multi-level and multi-granularity feature extractor is constructed by utilizing a level feature fusion method;
a data set processing module: after a small sample data set is constructed by using a meta task divider, training is carried out by using a support set sample and a query set sample, and optimized parameter information of the feature extractor is obtained;
test set processing module: and (3) carrying out test set evaluation on the training data of the working condition A through a multi-level feature extractor, and carrying out fine adjustment on parameters of the feature extractor by utilizing the training data of the working condition B, so as to evaluate the training data on the test set to achieve generalization of the fault automatic diagnosis method among different data distributions.
5. An air conditioner comprising a fan failure diagnosis system of an air conditioning system under weak supervision, which performs the fan failure diagnosis method of an air conditioning system under weak supervision as set forth in any one of claims 1 to 3 when operated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311714348.2A CN117404765B (en) | 2023-12-14 | 2023-12-14 | Air conditioner system fan fault diagnosis method and system under weak supervision condition and air conditioner |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311714348.2A CN117404765B (en) | 2023-12-14 | 2023-12-14 | Air conditioner system fan fault diagnosis method and system under weak supervision condition and air conditioner |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117404765A CN117404765A (en) | 2024-01-16 |
| CN117404765B true CN117404765B (en) | 2024-03-22 |
Family
ID=89492902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311714348.2A Active CN117404765B (en) | 2023-12-14 | 2023-12-14 | Air conditioner system fan fault diagnosis method and system under weak supervision condition and air conditioner |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117404765B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112418013A (en) * | 2020-11-09 | 2021-02-26 | 贵州大学 | Complex working condition bearing fault diagnosis method based on meta-learning under small sample |
| CN113486868A (en) * | 2021-09-07 | 2021-10-08 | 中南大学 | Motor fault diagnosis method and system |
| CN115859077A (en) * | 2022-11-10 | 2023-03-28 | 浙江舜云互联技术有限公司 | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions |
| CN116109292A (en) * | 2023-02-27 | 2023-05-12 | 重庆邮电大学 | Fan gear box fault diagnosis method based on federal semi-supervised learning |
| CN116451150A (en) * | 2023-04-04 | 2023-07-18 | 河北工业大学 | Equipment fault diagnosis method based on semi-supervised small sample |
| CN116702076A (en) * | 2023-05-25 | 2023-09-05 | 东北林业大学 | Small sample migration learning fault diagnosis method, system, computer and storage medium based on CNN feature fusion |
| CN117112992A (en) * | 2023-07-31 | 2023-11-24 | 东华大学 | Fault diagnosis method for polyester esterification stage |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334740A (en) * | 2019-06-05 | 2019-10-15 | 武汉大学 | The electrical equipment fault of artificial intelligence reasoning fusion detects localization method |
-
2023
- 2023-12-14 CN CN202311714348.2A patent/CN117404765B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112418013A (en) * | 2020-11-09 | 2021-02-26 | 贵州大学 | Complex working condition bearing fault diagnosis method based on meta-learning under small sample |
| CN113486868A (en) * | 2021-09-07 | 2021-10-08 | 中南大学 | Motor fault diagnosis method and system |
| CN115859077A (en) * | 2022-11-10 | 2023-03-28 | 浙江舜云互联技术有限公司 | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions |
| CN116109292A (en) * | 2023-02-27 | 2023-05-12 | 重庆邮电大学 | Fan gear box fault diagnosis method based on federal semi-supervised learning |
| CN116451150A (en) * | 2023-04-04 | 2023-07-18 | 河北工业大学 | Equipment fault diagnosis method based on semi-supervised small sample |
| CN116702076A (en) * | 2023-05-25 | 2023-09-05 | 东北林业大学 | Small sample migration learning fault diagnosis method, system, computer and storage medium based on CNN feature fusion |
| CN117112992A (en) * | 2023-07-31 | 2023-11-24 | 东华大学 | Fault diagnosis method for polyester esterification stage |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117404765A (en) | 2024-01-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112148955B (en) | Method and system for detecting abnormality of time sequence data of Internet of things | |
| CN109297689B (en) | Large-scale hydraulic machinery intelligent diagnosis method introducing weight factors | |
| CN112529234A (en) | Surface water quality prediction method based on deep learning | |
| CN116502072B (en) | Robust fault diagnosis method for key components of wind generating set under complex variable working conditions | |
| Bhakta et al. | Fault diagnosis of induction motor bearing using cepstrum-based preprocessing and ensemble learning algorithm | |
| CN116935892A (en) | Industrial valve anomaly detection method based on audio key feature dynamic aggregation | |
| DeVol et al. | Inception based deep convolutional neural network for remaining useful life estimation of turbofan engines | |
| CN115470717A (en) | Method, device, equipment and storage medium for predicting remaining life of robot | |
| CN117404765B (en) | Air conditioner system fan fault diagnosis method and system under weak supervision condition and air conditioner | |
| CN114118162A (en) | Bearing fault detection method based on improved deep forest algorithm | |
| CN116777109A (en) | Wind generating set equipment state evaluation method and system | |
| CN111504673A (en) | Fault diagnosis method and system for water chilling unit and air conditioner | |
| CN116085245A (en) | Online compressor performance prediction method and system based on OS-ELM | |
| CN113869194B (en) | Variable parameter milling process signal marking method and system based on deep learning | |
| CN112434729B (en) | Intelligent fault diagnosis method based on layer regeneration network under unbalanced sample | |
| CN114970950A (en) | Fan fault alarm method and device, storage medium and electronic equipment | |
| CN115184734A (en) | Power grid line fault detection method and system | |
| CN110929849B (en) | Video detection method and device based on neural network model compression | |
| Revathi et al. | Weather Prediction (analysis) using Soft kind of Computing techniques | |
| CN118035873B (en) | Fault diagnosis method for parallel convolution improved triple network | |
| CN117892224B (en) | Mixed characteristic neural network signal processing hardware circuit with wake-up mechanism | |
| CN115514621B (en) | Fault monitoring method, electronic device and storage medium | |
| CN114997246B (en) | Cavitation state identification method driven by fluid mechanical vibration data | |
| CN114199992B (en) | Method and system for detecting corrosion of tank wall of oil storage tank | |
| CN117909813A (en) | System for classifying and storing data by using deep learning technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |