CN116502134A - Self-diagnosis early warning abnormal functional state identification system - Google Patents
Self-diagnosis early warning abnormal functional state identification system Download PDFInfo
- Publication number
- CN116502134A CN116502134A CN202310251453.0A CN202310251453A CN116502134A CN 116502134 A CN116502134 A CN 116502134A CN 202310251453 A CN202310251453 A CN 202310251453A CN 116502134 A CN116502134 A CN 116502134A
- Authority
- CN
- China
- Prior art keywords
- data
- classifier
- self
- fault
- class
- 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
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 54
- 238000004092 self-diagnosis Methods 0.000 title claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 45
- 238000012545 processing Methods 0.000 claims abstract description 19
- 238000003745 diagnosis Methods 0.000 claims abstract description 10
- 230000005856 abnormality Effects 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000011156 evaluation Methods 0.000 claims description 33
- 238000000034 method Methods 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000013507 mapping Methods 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 8
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000002790 cross-validation Methods 0.000 claims description 6
- 238000013500 data storage Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 3
- 230000007488 abnormal function Effects 0.000 claims 1
- 238000012423 maintenance Methods 0.000 abstract description 5
- 230000009286 beneficial effect Effects 0.000 description 7
- 239000000463 material Substances 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000005243 fluidization Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G53/00—Conveying materials in bulk through troughs, pipes or tubes by floating the materials or by flow of gas, liquid or foam
- B65G53/34—Details
- B65G53/66—Use of indicator or control devices, e.g. for controlling gas pressure, for controlling proportions of material and gas, for indicating or preventing jamming of material
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Mechanical Engineering (AREA)
- Probability & Statistics with Applications (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention provides a self-diagnosis early warning abnormal functional state identification system, which relates to the technical field of thermal power plant safety and comprises a data acquisition module, a data acquisition module and a data processing module, wherein the data acquisition module is used for acquiring original data in real time; the data processing module is used for preprocessing the original data and the fault data set extracted from the historical data; the model building module is used for obtaining an abnormal self-diagnosis prediction model from a union set of the classifier obtained by training the processed data; the state identification module is used for inputting the data into the abnormal self-diagnosis prediction model, identifying fault data and sending corresponding alarm signals to the operator station based on the output first fault type. Obtaining an anomaly diagnosis prediction model through a union set of the classifier obtained by training the preprocessed fault data; based on the fault type output by the real-time data input abnormality diagnosis prediction model, corresponding alarm signals are determined and fed back to the operator station, so that the safe operation of the system is ensured, and meanwhile, a basis is provided for state maintenance.
Description
Technical Field
The invention relates to the technical field of safety of thermal power plants, in particular to a self-diagnosis early warning abnormal functional state identification system.
Background
The ash conveying system is an important component part in the dedusting system of the thermal power plant, the operation efficiency of the ash conveying system directly influences the dedusting efficiency and the operation state of the dedusting system, and along with the increasing importance of environmental protection, the requirement on dedusting of the thermal power plant is also higher and higher, and the operation of the ash conveying system is also very critical.
At present, the control of the ash conveying system still basically relies on preset parameters to automatically control ash conveying, and when pipe blockage, overtime conveying, pressurized leakage, valve blocking and abnormal material level occur in the ash conveying process, the failure judgment can not be timely found and accurately carried out, so that the conveying efficiency of the ash conveying system is reduced, and even equipment of the ash conveying system is damaged under serious conditions.
Therefore, the invention provides a self-diagnosis early warning abnormal functional state identification system.
Disclosure of Invention
The invention provides a self-diagnosis early warning function abnormal state identification system, which is used for obtaining an abnormal diagnosis prediction model by merging classifiers obtained by training preprocessed historical fault data; based on the fault type output by the real-time data input abnormality diagnosis prediction model, corresponding alarm signals are determined and fed back to the operator station, so that the safe operation of the system is ensured, and meanwhile, a basis is provided for state maintenance.
The invention provides a self-diagnosis early warning abnormal functional state identification system, which comprises:
and a data acquisition module: the device is used for acquiring original data in real time by using a preset device arranged at a preset position and storing the original data;
and a data processing module: the method comprises the steps of preprocessing original data and fault data sets extracted from historical data respectively to obtain first data and first target data;
and a model building module: the classifier is used for training based on the first target data, and the classifier is combined to obtain an abnormal self-diagnosis prediction model;
a state identification module: the first data are input into an abnormal self-diagnosis prediction model, fault data are identified, corresponding first fault categories are output, and alarm signals of the corresponding categories are sent to the operator station according to the first fault categories and fed back to the operator station.
Preferably, the data acquisition module includes:
device mounting unit: the system is used for installing a first pressure sensor at each ash conveying pipeline in an ash conveying system of a thermal power plant, respectively installing a second pressure sensor and a position feedback detection switch on a pneumatic control valve in front of a discharge valve of a conveying bin pump group and installing a passive nucleon level gauge at the lower position of an ash bucket;
a data acquisition unit: the system is used for respectively acquiring real-time data of the first pressure sensor, the second pressure sensor, the position feedback detection switch and the passive nucleon level gauge, integrating the corresponding time data and position data to obtain original data, and transmitting the original data to the data storage unit for storage.
Preferably, the data processing module is configured to perform redundancy processing on the original data and the fault data set, and then perform normalization processing on the original data and the fault data set to obtain first data and first target data.
Preferably, the model building module includes:
sample determination unit: for arbitrarily dividing the first target data into m groups, and determining the arbitrary m-1 groups of first target data as a training sample set y= { Y 1 ,y 2 ,y 3 ,…y n -wherein n is expressed as the total number of samples;
a neighbor domain construction unit: for utilizing the training sample set y= { Y 1 ,y 2 ,y 3 ,…y n Acquiring principal component information z= { Z } 1 ,z 2 ,z 3 ,…z n };
Based on Euclidean distance, constructing neighbor domain Z of low-dimensional space sample point i ={z i1 ,z i2 ,z i3 ,…z ik+1 (i=1, 2,3, …, n) and neighbor Y to the data space sample point i ={y i1 ,y i2 ,y i3 ,…y ik }(i=1,2,3,…,n);
Reconstructing a sample unit: the method comprises the steps of obtaining an affine matrix, and obtaining a reconstruction error based on local error inverse mapping of each low-dimensional sample point and the neighborhood center of the affine matrix;
alignment of each near neighborhood Y in original space i Center of the machineThe reconstruction error of (2) to obtain a reconstructed sample point y ri Finally, a reconstructed sample set Y is obtained r ={y r1 ,y r2 ,y r3 ,…y rn };
Model construction unit: the method is used for constructing a second classifier according to the reconstructed sample set, retraining to obtain an optimal second classifier based on parameters of the second classifier obtained by using a cross-checking algorithm, and merging to obtain an abnormal self-diagnosis prediction model.
Preferably, the model construction unit includes:
dividing blocks: for grouping the reconstructed sample set Y according to N kinds of labels of the reconstructed sample set r ={y r1 ,y r2 ,y r3 ,…y rn Dividing samples with different labels in every two in the sequence to construct the sequenceA second class of sample sets, wherein []Representing a rounding symbol;
training block: after training a classifier for each class-II sample set, sequentially inputting the rest group of first target data as data to be tested to judge classification results;
if only one classifier is found, marking the classifier as a first positive class, and finally classifying the classifier as a class label corresponding to the first positive class;
if more than one classifier results, selecting a class label corresponding to the classifier with highest reliability as a final classification result;
building block: the method is used for training again based on a cross-validation algorithm to obtain an optimal two-classifier, and then merging to obtain an abnormal self-diagnosis prediction model.
Preferably, the building block comprises:
parameter calculation unit: the method comprises the steps of obtaining first evaluation parameters of two classifiers based on judgment of classification results of data to be detected and combining with real classification results;
the first evaluation parameter is calculated as follows:
wherein P is j A first evaluation parameter denoted as a j-th classifier; y is jT The j-th classifier judges the true class of the data to be detected to the correct sample number; n represents the total number of data to be measured; y is jF The number of samples of the true class judgment errors of the data to be detected is represented as the j-th classifier;the precision expressed as j-th two classifiers; />The influence weight coefficient is expressed as the precision ratio on the first evaluation parameter; />A recall represented as the j-th classifier; />The influence weight expressed as recall ratio on the first evaluation parameterA weight coefficient; epsilon is expressed as a calculation loss factor for obtaining a first evaluation parameter;
if the first evaluation parameters of the two classifiers are smaller than a preset evaluation threshold, taking any one group of data in the first target data divided into m groups as data to be tested each time, taking the rest m-1 groups of data as a training sample set, and training to obtain a plurality of groups of corresponding to-be-determined two classifiers;
acquiring a first evaluation parameter of each group of two classifiers to be determined, and screening to obtain an optimal two classifier group with optimal classification effect by combining a preset evaluation threshold;
and merging all the two classifiers in the optimal two-classifier group to obtain an abnormal self-diagnosis prediction model.
Preferably, the method further comprises:
the calculation module: the reliability of the two classifiers is calculated, and the calculation formula is as follows:
wherein K is s Expressed as the reliability of the s-th classifier; g (·) is represented as a mapping function based on identifying nearest neighbor distance variables within the class and presenting increments, and has a value range of [0,1];l s (C a I x) represents that the s-th classifier recognizes the input data x to be measured as a class C a Identifying nearest neighbor distances within the class of decision outputs of (a); x is the data to be measured which is classified and judged by using the s-th classifier; c (C) a The class is expressed as a class obtained after the s-th classifier performs classification recognition on the data x to be detected, wherein a epsilon {1,2, …, r }; d, d s (C a |x) is expressed as a posterior probability criterion; h (·) is represented as a mapping function based on posterior probability;
when more than one classifier result exists, calculating the first reliability corresponding to each classifier result by using a formula;
and determining the class label corresponding to the corresponding two classifiers with the highest first reliability as a final classification result.
Preferably, the state recognition module includes:
an identification unit: the method comprises the steps of inputting first data into an abnormality diagnosis prediction model to output a first result, determining an output sample set as fault data according to the first result, and taking an output fault class as a first fault class;
and a feedback unit: and the alarm device is used for determining a corresponding alarm signal according to the first fault category and sending the alarm signal to an operator station.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
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. In the drawings:
fig. 1 is a block diagram of a system for identifying a self-diagnosis early warning abnormal functional state in an embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.
An embodiment of the present invention provides a self-diagnosis early warning abnormal functional state recognition system, as shown in fig. 1, including:
and a data acquisition module: the device is used for acquiring original data in real time by using a preset device arranged at a preset position and storing the original data;
and a data processing module: the method comprises the steps of preprocessing original data and fault data sets extracted from historical data respectively to obtain first data and first target data;
and a model building module: the classifier is used for training based on the first target data, and the classifier is combined to obtain an abnormal self-diagnosis prediction model;
a state identification module: the first data are input into an abnormal self-diagnosis prediction model, fault data are identified, corresponding first fault categories are output, and alarm signals of the corresponding categories are sent to the operator station according to the first fault categories and fed back to the operator station.
In the embodiment, the preset position refers to the position of each ash conveying pipeline in the ash conveying system of the thermal power plant, the front part of a discharge valve of a conveying bin pump group, a pneumatic control valve and the lower position of an ash bucket; the preset device comprises a first pressure sensor, a second pressure sensor, a position feedback detection switch and a passive nucleon level gauge.
In the embodiment, the original data comprises pressure data respectively collected in real time by a first pressure sensor and a second pressure sensor, position feedback signals collected in real time by a position feedback detection switch, material level data collected in real time by a passive nucleon level meter, current moment data of the data collected in real time and position data of all preset devices; the fault data set is composed of a certain preset amount of abnormal data extracted from the historical data.
In this embodiment, the first data and the first target data are obtained by performing redundancy processing on the original data and the failure data set, and then performing standardization processing, where the purpose of the redundancy processing is to reduce useless data and avoid resource waste; the purpose of the normalization process is to eliminate the influence of the dimensions of different data in the dataset on the data and reduce the distribution variance.
In the embodiment, the two classifiers are models which are trained based on first target data and used for classifying tasks of the two classes; the abnormal self-diagnosis prediction model is based on parameters of two classifiers and is obtained by integrating all the two classifiers and is used for identifying fault data and outputting corresponding fault categories.
In the embodiment, the fault data refers to data which is obtained in real time and is input into an abnormal self-diagnosis prediction model after being processed; the first fault category refers to a category which is obtained by inputting the original data acquired in real time and processed into an abnormal self-diagnosis prediction model and is output, such as pipe blockage, overtime conveying or abnormal material level; the operator station is operable to respond effectively in response to the acquired alarm signal.
The beneficial effects of the technical scheme are as follows: obtaining an anomaly diagnosis prediction model through the union of the classifiers obtained by training the preprocessed historical fault data; based on the fault type output by the real-time data input abnormality diagnosis prediction model, corresponding alarm signals are determined and fed back to the operator station, so that the safe operation of the system is ensured, and meanwhile, a basis is provided for state maintenance.
The embodiment of the invention provides a self-diagnosis early warning abnormal functional state identification system, wherein the data acquisition module comprises:
device mounting unit: the system is used for installing a first pressure sensor at each ash conveying pipeline in an ash conveying system of a thermal power plant, respectively installing a second pressure sensor and a position feedback detection switch on a pneumatic control valve in front of a discharge valve of a conveying bin pump group and installing a passive nucleon level gauge at the lower position of an ash bucket;
a data acquisition unit: the system is used for respectively acquiring real-time data of the first pressure sensor, the second pressure sensor, the position feedback detection switch and the passive nucleon level gauge, integrating the corresponding time data and position data to obtain original data, and transmitting the original data to the data storage unit for storage.
In the embodiment, the ash conveying system of the thermal power plant mainly comprises a bin type conveying pump, an ash conveying pipeline, an air source, an ash bucket and a control part; the first pressure sensor is used for measuring pressure data of the conveying pipeline in real time so as to judge whether overtime conveying and pipe blockage exist or not; the second pressure sensor is used for measuring the pressure value of the ash conveying bin pump group in real time and corresponds to conveying tightness in the stamping fluidization stage so as to judge whether pressurizing leakage exists or not; the passive nuclear level gauge is used for monitoring and controlling the level in the bin pump to judge whether the level is abnormal or not; the position feedback detection switch is used for collecting position feedback signals in real time so as to judge whether a valve jam fault condition exists.
In this embodiment, the original data is composed of data collected in real time by all preset devices, corresponding time data and position data, wherein the time data refers to the current time data of the preset devices for collecting the data in real time; the position data refers to a specific location where the preset device is installed.
The beneficial effects of the technical scheme are as follows: real-time data are collected by using a pressure sensor, a position feedback switch device and a passive nucleon level gauge which are arranged at a preset position, and then the real-time data are transmitted to a data storage unit for storage, so that data support is provided for subsequent fault identification.
The embodiment of the invention provides a self-diagnosis early warning abnormal functional state identification system, wherein a model building module comprises:
sample determination unit: for arbitrarily dividing the first target data into m groups, and determining the arbitrary m-1 groups of first target data as a training sample set y= { Y 1 ,y 2 ,y 3 ,…y n -wherein n is expressed as the total number of samples;
a neighbor domain construction unit: for utilizing the training sample set y= { Y 1 ,y 2 ,y 3 ,…y n Acquiring principal component information z= { Z } 1 ,z 2 ,z 3 ,…z n };
Based on Euclidean distance, constructing neighbor domain Z of low-dimensional space sample point i ={z i1 ,z i2 ,z i3 ,…z ik+1 (i=1, 2,3, …, n) and neighbor Y to the data space sample point i ={y i1 ,y i2 ,y i3 ,…y ik }(i=1,2,3,…,n);
Reconstructing a sample unit: the method comprises the steps of obtaining an affine matrix, and obtaining a reconstruction error based on local error inverse mapping of each low-dimensional sample point and the neighborhood center of the affine matrix;
alignment of each near neighborhood Y in original space i Center of the machineReconstruction errors of (a)Difference, obtain reconstructed sample point y ri Finally, a reconstructed sample set Y is obtained r ={y r1 ,y r2 ,y r3 ,…y rn };
Model construction unit: the method is used for constructing a second classifier according to the reconstructed sample set, retraining to obtain an optimal second classifier based on parameters of the second classifier obtained by using a cross-checking algorithm, and merging to obtain an abnormal self-diagnosis prediction model.
In this embodiment, the first target data is data obtained by performing redundancy processing on the failure data set and then performing standardization processing; the training sample set is composed of any m-1 set of first target data.
In the embodiment, the purpose of acquiring the principal component information of the training sample set is to extract the data global feature information, reduce the complexity of the data features while ensuring the validity of the data information, and improve the subsequent training effect on the model; the purpose of the affine matrix is to inversely map local structure errors to high-dimensional space, where local errors refer to the difference between each low-dimensional sample point and its neighborhood center; the aim of reconstructing errors of the centers of each near neighborhood in the original space is to acquire lost local structure information, so that the data expression capacity of the sample is improved, wherein the reconstruction errors are obtained by inverse mapping of the local errors.
In this embodiment, the two classifiers are models that perform classification tasks for two classes; the cross checking algorithm is mainly used for acquiring parameters of the two classifiers obtained by inputting different groups of data to be tested into different groups of training sample sets for training, so that the classification effect of the two classifiers corresponding to the parameters is relatively objectively judged; the abnormal self-diagnosis prediction model is obtained by combining classifiers obtained through final training and is used for identifying fault data and fault types in real time.
The beneficial effects of the technical scheme are as follows: the method comprises the steps of taking most of first target data as training samples, obtaining principal component information, constructing a near neighborhood for each coordinate in a low-dimensional space, obtaining a local error between each coordinate and the center of the neighborhood, reconstructing the error back to the high-dimensional space by inverse mapping, and training the reconstructed samples by using the obtained reconstructed coordinates as reconstructed samples to obtain a classifier, so that an abnormal self-diagnosis prediction model is constructed, the expression capacity of the data is improved, the classification effect of the classifier is ensured, and the accurate recognition function of the abnormal self-diagnosis prediction model is guaranteed.
The embodiment of the invention provides a self-diagnosis early warning abnormal functional state identification system, wherein a model building unit comprises:
dividing blocks: for grouping the reconstructed sample set Y according to N kinds of labels of the reconstructed sample set r ={y r1 ,y r2 ,y r3 ,…y rn Dividing samples with different labels in every two in the sequence to construct the sequenceA second class of sample sets, wherein []Representing a rounding symbol;
training block: after training a classifier for each class-II sample set, sequentially inputting the rest group of first target data as data to be tested to judge classification results;
if only one classifier is found, marking the classifier as a first positive class, and finally classifying the classifier as a class label corresponding to the first positive class;
if more than one classifier results, selecting a class label corresponding to the classifier with highest reliability as a final classification result;
building block: the method is used for training again based on a cross-validation algorithm to obtain an optimal two-classifier, and then merging to obtain an abnormal self-diagnosis prediction model.
In this embodiment, the class label refers to a data classification name, such as a pump group pressure value of an ash conveying bin and a pressure value of a conveying pipeline; the two-class sample set is composed of two different classes of samples.
In this embodiment, for example, there are first target data divided into 10 groups, of which any 9 groups of data are used as training sample sets, and the remaining group of first target data are used as data to be measured.
In this embodiment, for example, there is data D1 to be measured for classification, two results of the classifiers C1 and C2 are obtained, at this time, the reliability of the classifiers C1 and C2 is obtained respectively, and a class label corresponding to the classifier with the highest reliability is selected as the final classification result.
In the embodiment, the cross-validation algorithm is mainly used for acquiring parameters of a classifier obtained by inputting different groups of data to be tested into different groups of training sample sets for training; the optimal two classifiers are obtained by re-acquiring proper training data based on a cross-validation algorithm and training; and obtaining the abnormal self-diagnosis prediction model union optimal two classifiers.
The beneficial effects of the technical scheme are as follows: training the two classifiers by utilizing the reconstructed sample set, and obtaining corresponding classification results based on the data to be tested; based on the effect analysis of classification results, the optimal two classifiers are obtained through retraining by combining a cross-validation algorithm, and then the abnormal self-diagnosis prediction model is built by union, so that the accurate recognition function of the abnormal self-diagnosis prediction model is ensured.
The embodiment of the invention provides a self-diagnosis early warning abnormal functional state identification system, which comprises the following building blocks:
parameter calculation unit: the method comprises the steps of obtaining first evaluation parameters of two classifiers based on judgment of classification results of data to be detected and combining with real classification results;
the first evaluation parameter is calculated as follows:
wherein P is j A first evaluation parameter denoted as a j-th classifier; y is jT The j-th classifier judges the true class of the data to be detected to the correct sample number; n represents the total number of data to be measured; y is jF The number of samples of the true class judgment errors of the data to be detected is represented as the j-th classifier;the precision expressed as j-th two classifiers; />The influence weight coefficient is expressed as the precision ratio on the first evaluation parameter; />A recall represented as the j-th classifier; />The influence weight coefficient is expressed as the recall ratio on the first evaluation parameter; epsilon is expressed as a calculation loss factor for obtaining a first evaluation parameter;
if the first evaluation parameters of the two classifiers are smaller than a preset evaluation threshold, taking any one group of data in the first target data divided into m groups as data to be tested each time, taking the rest m-1 groups of data as a training sample set, and training to obtain a plurality of groups of corresponding to-be-determined two classifiers;
acquiring a first evaluation parameter of each group of two classifiers to be determined, and screening to obtain an optimal two classifier group with optimal classification effect by combining a preset evaluation threshold;
and merging all the two classifiers in the optimal two-classifier group to obtain an abnormal self-diagnosis prediction model.
In this embodiment, the preset evaluation threshold is set in advance, and is generally 0.85; the undetermined two classifiers are classifiers obtained by training based on taking any one group of data in first target data divided into m groups as data to be tested in sequence and the rest m-1 groups of data as training sample sets; the optimal two classifier groups are composed of classifiers with optimal classification effects, namely the classifiers with optimal first evaluation parameters, and are used for obtaining an abnormal self-diagnosis prediction model by union.
The beneficial effects of the technical scheme are as follows: judging the classification effect of the two classifiers by acquiring first evaluation parameters of the two classifiers; and combining a cross verification algorithm, retraining the obtained optimal two classifiers by using the most suitable training data, and collecting to obtain an abnormal self-diagnosis prediction model with optimal recognition function, so that the fault recognition efficiency is improved, the safe operation of the system is ensured, and meanwhile, an accurate basis is provided for state maintenance.
The embodiment of the invention provides a self-diagnosis early warning abnormal functional state identification system, which further comprises:
the calculation module: the reliability of the two classifiers is calculated, and the calculation formula is as follows:
wherein K is s Expressed as the reliability of the s-th classifier; g (·) is represented as a mapping function based on identifying nearest neighbor distance variables within the class and presenting increments, and has a value range of [0,1];l s (C a I x) represents that the s-th classifier recognizes the input data x to be measured as a class C a Identifying nearest neighbor distances within the class of decision outputs of (a); x is the data to be measured which is classified and judged by using the s-th classifier; c (C) a The class is expressed as a class obtained after the s-th classifier performs classification recognition on the data x to be detected, wherein a epsilon {1,2, …, r }; d, d s (C a |x) is expressed as a posterior probability criterion; h (·) is represented as a mapping function based on posterior probability;
when more than one classifier result exists, calculating the first reliability corresponding to each classifier result by using a formula;
and determining the class label corresponding to the corresponding two classifiers with the highest first reliability as a final classification result.
In this embodiment, for example, there is a corresponding reliability K of acquiring the two classifier results c1, c2, c3 1 >K 2 =K 3 The method comprises the steps of carrying out a first treatment on the surface of the At this time, the class label corresponding to the two classifier results c1 is used as the final classification result.
The beneficial effects of the technical scheme are as follows: through the reliability of the classifier based on the calculation, when more than one classifier result exists, the class label corresponding to the classifier with the highest reliability is determined to be output as the first fault class, so that the accuracy of providing basis for state maintenance is ensured, and the safe operation of the system is indirectly ensured.
The embodiment of the invention provides a self-diagnosis early warning function abnormal state identification system, wherein a state identification module comprises:
an identification unit: the method comprises the steps of inputting first data into an abnormality diagnosis prediction model to output a first result, determining an output sample set as fault data according to the first result, and taking an output fault class as a first fault class;
and a feedback unit: and the alarm device is used for determining a corresponding alarm signal according to the first fault category and sending the alarm signal to an operator station.
In this embodiment, the first data is obtained by performing redundancy processing on the original data and then performing normalization processing; the first result comprises an output sample set and a fault class; the alarm signal is a signal sent by a computer display device, an acoustic device or an indicator lamp and is used for reminding an emergency that the ash conveying system fails; the operator station can quickly learn the fault position and the fault reason according to the acquired different alarm signals, so that corresponding measures are taken.
The beneficial effects of the technical scheme are as follows: the fault data and the corresponding fault types are quickly obtained by inputting the preprocessed real-time collected original data into the abnormality diagnosis prediction model, and then the corresponding alarm signals are sent to the operator station according to the fault types, so that the error rate of manual judgment is avoided, the processing timeliness of the fault is improved, and the safe operation of the system is ensured.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (8)
1. A self-diagnosis early-warning abnormal-function state recognition system, characterized by comprising:
and a data acquisition module: the device is used for acquiring original data in real time by using a preset device arranged at a preset position and storing the original data;
and a data processing module: the method comprises the steps of preprocessing original data and fault data sets extracted from historical data respectively to obtain first data and first target data;
and a model building module: the classifier is used for training based on the first target data, and the classifier is combined to obtain an abnormal self-diagnosis prediction model;
a state identification module: the first data are input into an abnormal self-diagnosis prediction model, fault data are identified, corresponding first fault categories are output, and alarm signals of the corresponding categories are sent to the operator station according to the first fault categories and fed back to the operator station.
2. The self-diagnosis early warning abnormal state recognition system according to claim 1, wherein the data acquisition module comprises:
device mounting unit: the system is used for installing a first pressure sensor at each ash conveying pipeline in an ash conveying system of a thermal power plant, respectively installing a second pressure sensor and a position feedback detection switch on a pneumatic control valve in front of a discharge valve of a conveying bin pump group and installing a passive nucleon level gauge at the lower position of an ash bucket;
a data acquisition unit: the system is used for respectively acquiring real-time data of the first pressure sensor, the second pressure sensor, the position feedback detection switch and the passive nucleon level gauge, integrating the corresponding time data and position data to obtain original data, and transmitting the original data to the data storage unit for storage.
3. The system for identifying a self-diagnosis early warning abnormal functional state according to claim 1, wherein the data processing module is configured to perform redundancy processing on the original data and the fault data set respectively, and then perform normalization processing to obtain first data and first target data.
4. The system for identifying a self-diagnosis early warning abnormal functional state according to claim 1, wherein the model building module comprises:
sample determination unit:for arbitrarily dividing the first target data into m groups, and determining the arbitrary m-1 groups of first target data as a training sample set y= { Y 1 ,y 2 ,y 3 ,…y n -wherein n is expressed as the total number of samples;
a neighbor domain construction unit: for utilizing the training sample set y= { Y 1 ,y 2 ,y 3 ,…y n Acquiring principal component information z= { Z } 1 ,z 2 ,z 3 ,…z n };
Based on Euclidean distance, constructing neighbor domain Z of low-dimensional space sample point i ={z i1 ,z i2 ,z i3 ,…z ik+1 (i=1, 2,3, …, n) and neighbor Y to the data space sample point i ={y i1 ,y i2 ,y i3 ,…y ik }(i=1,2,3,…,n);
Reconstructing a sample unit: the method comprises the steps of obtaining an affine matrix, and obtaining a reconstruction error based on local error inverse mapping of each low-dimensional sample point and the neighborhood center of the affine matrix;
alignment of each near neighborhood Y in original space i Center of the machineThe reconstruction error of (2) to obtain a reconstructed sample point y ri Finally, a reconstructed sample set Y is obtained r ={y r1 ,y r2 ,y r3 ,…y rn };
Model construction unit: the method is used for constructing a second classifier according to the reconstructed sample set, retraining to obtain an optimal second classifier based on parameters of the second classifier obtained by using a cross-checking algorithm, and merging to obtain an abnormal self-diagnosis prediction model.
5. The self-diagnosis early-warning abnormal-state recognition system according to claim 4, wherein the model construction unit includes:
dividing blocks: for grouping the reconstructed sample set Y according to N kinds of labels of the reconstructed sample set r ={y r1 ,y r2 ,y r3 ,…y rn Dividing samples with different labels in every two in the sequence to construct the sequenceA second class of sample sets, wherein []Representing a rounding symbol;
training block: after training a classifier for each class-II sample set, sequentially inputting the rest group of first target data as data to be tested to judge classification results;
if only one classifier is found, marking the classifier as a first positive class, and finally classifying the classifier as a class label corresponding to the first positive class;
if more than one classifier results, selecting a class label corresponding to the classifier with highest reliability as a final classification result;
building block: the method is used for training again based on a cross-validation algorithm to obtain an optimal two-classifier, and then merging to obtain an abnormal self-diagnosis prediction model.
6. The self-diagnosis early warning abnormal state recognition system according to claim 5, wherein the building block comprises:
parameter calculation unit: the method comprises the steps of obtaining first evaluation parameters of two classifiers based on judgment of classification results of data to be detected and combining with real classification results;
the first evaluation parameter is calculated as follows:
wherein P is j A first evaluation parameter denoted as a j-th classifier; y is jT The j-th classifier judges the true class of the data to be detected to the correct sample number; n represents the total number of data to be measured; y is jF The number of samples of the true class judgment errors of the data to be detected is represented as the j-th classifier;the precision expressed as j-th two classifiers; />The influence weight coefficient is expressed as the precision ratio on the first evaluation parameter; />A recall represented as the j-th classifier; />The influence weight coefficient is expressed as the recall ratio on the first evaluation parameter; epsilon is expressed as a calculation loss factor for obtaining a first evaluation parameter;
if the first evaluation parameters of the two classifiers are smaller than a preset evaluation threshold, taking any one group of data in the first target data divided into m groups as data to be tested each time, taking the rest m-1 groups of data as a training sample set, and training to obtain a plurality of groups of corresponding to-be-determined two classifiers;
acquiring a first evaluation parameter of each group of two classifiers to be determined, and screening to obtain an optimal two classifier group with optimal classification effect by combining a preset evaluation threshold;
and merging all the two classifiers in the optimal two-classifier group to obtain an abnormal self-diagnosis prediction model.
7. The self-diagnosis early-warning abnormal state recognition system according to claim 5, further comprising:
the calculation module: the reliability of the two classifiers is calculated, and the calculation formula is as follows:
wherein K is s Expressed as the reliability of the s-th classifier; g (·) is expressed as identifying nearest neighbor distance variable based on within classAnd presents an increasing mapping function with a value range of [0,1 ]];l s (C a I x) represents that the s-th classifier recognizes the input data x to be measured as a class C a Identifying nearest neighbor distances within the class of decision outputs of (a); x is the data to be measured which is classified and judged by using the s-th classifier; c (C) a The classification is expressed as a class obtained after the s-th classifier performs classification recognition on the data x to be detected, wherein a is epsilon {1,2,., r }; d, d s (C a |x) is expressed as a posterior probability criterion; h (·) is represented as a mapping function based on posterior probability;
when more than one classifier result exists, calculating the first reliability corresponding to each classifier result by using a formula;
and determining the class label corresponding to the corresponding two classifiers with the highest first reliability as a final classification result.
8. The self-diagnosis early warning abnormal state recognition system according to claim 1, wherein the state recognition module comprises:
an identification unit: the method comprises the steps of inputting first data into an abnormality diagnosis prediction model to output a first result, determining an output sample set as fault data according to the first result, and taking an output fault class as a first fault class;
and a feedback unit: and the alarm device is used for determining a corresponding alarm signal according to the first fault category and sending the alarm signal to an operator station.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310251453.0A CN116502134A (en) | 2023-03-15 | 2023-03-15 | Self-diagnosis early warning abnormal functional state identification system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310251453.0A CN116502134A (en) | 2023-03-15 | 2023-03-15 | Self-diagnosis early warning abnormal functional state identification system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116502134A true CN116502134A (en) | 2023-07-28 |
Family
ID=87319185
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310251453.0A Pending CN116502134A (en) | 2023-03-15 | 2023-03-15 | Self-diagnosis early warning abnormal functional state identification system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116502134A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117272179A (en) * | 2023-11-22 | 2023-12-22 | 南京迅集科技有限公司 | Equipment state detection system and method based on Internet of things |
-
2023
- 2023-03-15 CN CN202310251453.0A patent/CN116502134A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117272179A (en) * | 2023-11-22 | 2023-12-22 | 南京迅集科技有限公司 | Equipment state detection system and method based on Internet of things |
CN117272179B (en) * | 2023-11-22 | 2024-02-06 | 南京迅集科技有限公司 | Equipment state detection system and method based on Internet of things |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113255795B (en) | Equipment state monitoring method based on multi-index cluster analysis | |
US10977568B2 (en) | Information processing apparatus, diagnosis method, and program | |
US9060211B2 (en) | Method of collecting device-condition information and device-condition information collecting kit used therefor | |
CN112508105B (en) | Fault detection and retrieval method for oil extraction machine | |
US20160369777A1 (en) | System and method for detecting anomaly conditions of sensor attached devices | |
CN100461044C (en) | Melt index detection fault diagnozing system and method in propylene polymerization production | |
CN102799171A (en) | Detecting anomalies in fault code settings and enhancing service documents using analytical symptoms | |
CN104677997B (en) | A kind of transformer oil chromatographic on-line monitoring differentiation method for early warning | |
CN116502134A (en) | Self-diagnosis early warning abnormal functional state identification system | |
JP6674033B2 (en) | Plant abnormality diagnosis device and plant abnormality diagnosis system | |
CN109298633A (en) | Chemical production process fault monitoring method based on adaptive piecemeal Non-negative Matrix Factorization | |
US20210287515A1 (en) | Method for pre-detecting abnormality sign of nuclear power plant device including processor for determining device importance and warning validity, and system therefor | |
TWI752638B (en) | Method and system for detection of driving anomaly | |
CN108257365A (en) | A kind of industrial alarm designs method based on global nonspecific evidence dynamic fusion | |
CN108763966A (en) | A kind of Tail gas measuring cheating supervisory systems and method | |
CN206833239U (en) | A kind of thermal power plant's control system fault detection system based on data-driven | |
CN110455370B (en) | Flood-prevention drought-resisting remote measuring display system | |
CN100461043C (en) | Melt index detection fault diagnosis system and method for industial polypropylene production | |
CN114577470A (en) | Fault diagnosis method and system for fan main bearing | |
CN116610915A (en) | Pneumatic ash conveying abnormal state diagnosis and fault early warning method | |
CN104568438A (en) | Engine bearing fault detection system and method | |
CN114454671B (en) | Tire self-sealing emergency system based on tire pressure monitoring triggering | |
CN116204825A (en) | Production line equipment fault detection method based on data driving | |
CN206656749U (en) | Heat-engine plant sensor fault diagnosis system | |
EP3457325A1 (en) | Failure identification of a rail vehicle component |
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 |