CN114912372B - High-precision filling pipeline fault early warning method based on artificial intelligence algorithm - Google Patents
High-precision filling pipeline fault early warning method based on artificial intelligence algorithm Download PDFInfo
- Publication number
- CN114912372B CN114912372B CN202210683749.5A CN202210683749A CN114912372B CN 114912372 B CN114912372 B CN 114912372B CN 202210683749 A CN202210683749 A CN 202210683749A CN 114912372 B CN114912372 B CN 114912372B
- Authority
- CN
- China
- Prior art keywords
- pressure
- model
- data
- early warning
- sample
- 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
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 15
- 238000012544 monitoring process Methods 0.000 claims abstract description 35
- 238000007637 random forest analysis Methods 0.000 claims abstract description 33
- 239000002002 slurry Substances 0.000 claims description 19
- 238000012360 testing method Methods 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 16
- 238000003066 decision tree Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 230000005856 abnormality Effects 0.000 claims description 3
- 238000007499 fusion processing Methods 0.000 claims description 3
- 238000005381 potential energy Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 2
- 238000002156 mixing Methods 0.000 claims description 2
- 230000002159 abnormal effect Effects 0.000 abstract description 2
- 238000005065 mining Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003756 stirring Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Pipeline Systems (AREA)
Abstract
The invention discloses a high-precision filling pipeline fault early warning method based on an artificial intelligence algorithm, which comprises the following steps: collecting the concentration and flow data of a surface filling station, and obtaining the pressure and time data of key monitoring points of underground pipelines; establishing an improved random forest algorithm model and a fitting pressure function model; and judging the prediction precision of the improved random forest algorithm model and the fitted pressure function model, if the prediction precision meets the standard, constructing a pressure prediction model, predicting the pressure of key monitoring points and carrying out pipeline fault early warning. The method judges the working state of the filling pipeline by establishing a pressure prediction model, and realizes early warning of abnormal and fault states of the pipeline.
Description
Technical Field
The invention relates to the technical field of mine filling, in particular to a high-precision filling pipeline fault early warning method based on an artificial intelligence algorithm.
Background
The filling mining method is one of the most effective methods for ensuring the safety of deep mining, is also the best way for fully utilizing mine tailings, and gradually prolongs the filling slurry conveying pipeline, has more complex space arrangement structure, increases the fault points of the filling slurry conveying pipeline and causes faults frequently along with the increase of the mining depth. At present, the safety of a filling pipeline is mainly ensured by means of manual inspection, although a related automatic monitoring method for the fault of the filling pipeline is available, the existing monitoring method cannot achieve the effect of early warning in advance, judgment and reminding can only be carried out after the pipeline is blocked or leaked, and a manager cannot know the pressure change condition of the filling pipeline and can not timely make a control decision, so that the fault of the filling pipeline cannot be effectively avoided.
Disclosure of Invention
The invention provides a high-precision filling pipeline fault early warning method based on an artificial intelligence algorithm, which aims at: and judging the working state of the filling pipeline by establishing a pressure prediction model, so as to realize early warning of abnormal and fault states of the pipeline.
The technical scheme of the invention is as follows:
a high-precision filling pipeline fault early warning method based on an artificial intelligence algorithm comprises the following steps:
s1: collecting the concentration and flow data of a surface filling station, and obtaining the pressure and time data of key monitoring points of underground pipelines;
s2: establishing an improved random forest algorithm model and a fitting pressure function model;
s3: judging the prediction precision of the improved random forest algorithm model and the fitting pressure function model, if the prediction precision meets the standard, executing the step S4, otherwise, returning to the step S1;
s4: and constructing a pressure prediction model, predicting the pressure of the key monitoring point and carrying out pipeline fault early warning.
Further, the method for establishing the improved random forest algorithm model in the step S2 includes:
s21: normalizing the concentration, flow, monitoring point pressure and time data obtained in the step S1 to obtain an original data set;
s22: preprocessing an original data set in a mode of combining an improved SMOTE algorithm with mixed sampling to obtain a balanced data set;
s23: performing feature weighting and dimension reduction on the balanced data set through a ReLieff algorithm, screening out a subset training decision tree with a lower effective dimension, and obtaining an optimized data set by adopting a weighted voting method;
s24: dividing an optimized data set into a training set and a testing set, establishing an improved random forest algorithm model, using a square error as a model prediction precision standard, determining model parameters by using a contrast divergence training algorithm, normalizing the testing set by using a mean_train and a variance std_train obtained by normalizing the training set, checking model prediction precision by using the testing set, and judging the current model as the improved random forest algorithm model if the error between a model prediction pressure value and an actual measurement pressure value meets a prediction precision requirement.
Further, the step S22 specifically includes:
s22-1: clustering the minority class by using a clustering algorithm, determining the distribution state of each cluster in the minority class, then calculating cluster centers, and interpolating and adding sample data on a cluster center and intra-cluster sample connection line according to the following formula:
X new ＝C i +rand(0,1)*(X-C i )i＝1,2...N
wherein X is new To insert samples newly, C i Is a cluster center, X is C i The method comprises the steps of taking an original sample in cluster core clustering;
s22-2: sequentially calculating the sampling number of most types of samples and few types of samples:
P mix ＝Q-L
PM＝P mix ·t
PN＝P mix ·(1-t)
wherein Q is the number of most samples, L is the number of few samples, P mix PM is undersampled quantity, PN is oversampled quantity, and t is a mixing proportion coefficient;
s22-3: undersampling is carried out on most types of samples, oversampling is carried out on few types of samples, brand new data sets are respectively obtained, and the brand new data sets are combined to obtain a balanced data set D new 。
Further, the step S23 specifically includes:
s23-1: will balance data set D new The characteristic weight of each sample is set to 0;
s23-2: randomly selecting a sample R from the balanced data set samples, selecting k similar nearest neighbor samples of the sample R from similar sample sets of the sample R, respectively selecting k different types of nearest neighbor samples from different types of sample sets of each sample R, and calculating to obtain a characteristic weight corresponding to each type in the sample;
s23-3: repeating the step S23-2 for m times to obtain m feature weights of each class, calculating a feature weight average value of each class as a weight value W of each class, arranging the classes in descending order according to the weight values W, uniformly dividing the features into high, medium and low 3 areas according to the weight values, carrying out clustering reduction, and screening out a subset training decision tree with lower dimensionality through the obtained 3 feature areas;
s23-4: random sample balance dataset D new Obtaining a sample subset, recording an unextracted data set as an OT data set, obtaining the ratio of each decision tree to the correct classification of the OT data through the OT data set, and calculating the weight value of each decision tree, wherein the calculation formula is as follows:
wherein F is j Is a classification ratio;
and selecting the decision tree with the highest weight as a final classification result to generate an optimized data set.
Further, the method for establishing the fitting pressure function model in step S2 is as follows: establishing a monitoring point pressure, concentration and flow function relation by adopting a data fitting mode: p=f (Q, C V ) Fitting the functional relations of various structures, comparing standard errors and fitting goodness of the functional relations, and selecting a fitting pressure function according to the principles of small standard errors and large fitting goodness; after the fitting pressure function is selected, data fitting is carried out on different monitoring points respectively.
Further, the method for judging and improving the prediction precision of the random forest algorithm model and the fitting pressure function model in the step S3 is as follows: and carrying out mean square error calculation on the pressure predicted value of the improved random forest algorithm model and the pressure predicted value of the fitted pressure function model, and judging that the prediction precision reaches the standard if the calculated value is in a set range.
Further, the pressure prediction model in step S4 is: and the pressure predicted value of the improved random forest algorithm model and the pressure predicted value of the fitted pressure function model respectively occupy a certain weight to obtain the final pressure predicted value of the monitoring point.
Further, the step S1 further comprises preprocessing the acquired data, carrying out data fusion processing by taking the time difference of the ground surface slurry reaching the pressure monitoring point as a reference, and eliminating system abnormality and parking data.
Further, the method for acquiring pressure and time data of the key monitoring points of the underground pipeline in the step S1 comprises the following steps: determining a pressure value generated by potential energy of slurry at a lower port of a pipeline in a vertical section, subtracting a slurry along-path pressure loss value obtained by a loop test, and obtaining a slurry pressure value at any point of the pipeline; and taking the positions corresponding to the maximum pressure value and the minimum pressure value as key point positions of the underground filling pipeline, installing a pressure transmitter at the key point positions to measure pressure data, and recording the pressure and time data by using a paperless recorder.
Compared with the prior art, the invention has the following beneficial effects:
(1) The method fully utilizes the concentration, flow and pressure data of mine filling to establish an improved random forest algorithm pressure prediction model and a fitting pressure function relation model at key monitoring points of the filling pipeline, comprehensively predicts the pressure value of slurry when the slurry reaches the pressure monitoring points by using two pressure calculation models and carries out fault early warning, has high pressure prediction accuracy and strong implementation, and can be directly applied to the existing automatic filling system without newly adding complex pipeline state monitoring equipment;
(2) The method adopts an improved SMOTE algorithm and a mixed sampling method to improve random forest algorithm model parameters, preprocesses an original high-dimensional unbalanced data set, improves the balance degree of the data set, further improves the classification performance of an improved random forest algorithm model through a ReLieff algorithm and a weighted voting principle, ensures the fitting goodness of the improved random forest algorithm model, and improves the accuracy rate of pressure prediction;
(3) According to the method, a data fitting mode is adopted to establish functional relations of various structures, the functional relations with small standard error and large fitting goodness are screened out to serve as fitting pressure functions, data fitting is respectively carried out on different monitoring points, and accuracy of pressure prediction is further improved.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is an error diagram of a pressure prediction model of an improved random forest algorithm;
FIG. 3 is a schematic diagram of a fitted pressure function model.
Detailed Description
The technical scheme of the invention is described in detail below with reference to the accompanying drawings:
referring to fig. 1, a high-precision filling pipeline fault early warning method based on an artificial intelligence algorithm comprises the following steps:
s1: and collecting the concentration and flow data of the surface filling station, and obtaining the pressure and time data of the key monitoring points of the underground pipeline.
Preferably, when the filling station performs filling operation, different data acquisition software is adopted for different filling upper computer software to read the concentration, flow data and filling time data of the filling automation server, such as: the filling automation system upper computer adopts WinCC software, and Siemens WinCC OLE DB Provider software can be adopted for data extraction.
The method for acquiring the pressure and time data of the key monitoring points of the underground pipeline comprises the following steps: firstly, according to the principle of conservation of energy, the pressure distribution of a filling pipeline is obtained through preliminary calculation, specifically, the pressure value generated by the potential energy of slurry at the lower opening of a pipeline at the vertical section is determined, the slurry pressure value at any point of the pipeline can be obtained by subtracting the slurry along-line pressure loss value obtained by a loop test, the positions corresponding to the maximum pressure value and the minimum pressure value are used as key point positions of the underground filling pipeline, then, pressure transmitters are arranged at the key point positions to measure pressure data, and a paperless recorder is used for recording pressure and time data.
Further, preprocessing the acquired data, carrying out data fusion processing by taking the time difference of the ground surface slurry reaching the pressure monitoring point as a reference, and eliminating system abnormality and parking data.
S2: and establishing an improved random forest algorithm model and a fitting pressure function model.
Preferably, the method for establishing the improved random forest algorithm model comprises the following steps:
s21: and (3) carrying out normalization processing on the concentration, flow, monitoring point pressure and time data obtained in the step (S1) to obtain an original data set. The normalization formula is as follows:
wherein x is a characteristic value, x mean And S is the mean value corresponding to the characteristic value, and S is the variance corresponding to the characteristic value.
S22: the original data set is preprocessed in a mode of combining an improved SMOTE algorithm and mixed sampling to obtain a balanced data set, and the model fitting goodness is guaranteed to be more than 0.97.
The method specifically comprises the following steps:
s22-1: clustering the minority class by using a clustering algorithm, determining the distribution state of each cluster in the minority class, then calculating cluster centers, and interpolating and adding sample data on a cluster center and intra-cluster sample connection line according to the following formula:
X new ＝C i +rand(0,1)*(X-C i )i＝1,2...N
wherein X is new To insert samples newly, C i Is a cluster center, X is C i Is the original sample in the cluster core cluster.
S22-2: sequentially calculating the sampling number of most types of samples and few types of samples:
P mix ＝Q-L
PM＝P mix ·t
PN＝P mix ·(1-t)
wherein Q is the number of most samples, L is the number of few samples, P mix For the mixed sample number, PM is the undersampled number, PN is the oversampled number, and t is the mixed scaling factor.
S22-3: undersampling is carried out on most types of samples, oversampling is carried out on few types of samples, brand new data sets are respectively obtained, and the brand new data sets are combined to obtain a balanced data set D new 。
S23: and carrying out feature weighting and dimension reduction on the balanced data set through a ReLieff algorithm, screening out an effective subset training decision tree with lower dimension, and obtaining an optimized data set by adopting a weighted voting method.
The method specifically comprises the following steps:
s23-1: will balance data set D new The characteristic weight of each sample is set to 0;
s23-2: randomly selecting a sample R from the balanced data set samples, selecting k similar nearest neighbor samples of the sample R from similar sample sets of the sample R, respectively selecting k different types of nearest neighbor samples from different types of sample sets of each sample R, and calculating to obtain a characteristic weight corresponding to each type in the sample;
s23-3: repeating the step S23-2 for m times to obtain m feature weights of each class, calculating a feature weight average value of each class as a weight value W of each class, arranging the classes in descending order according to the weight values W, uniformly dividing the features into high, medium and low 3 areas according to the weight values, carrying out clustering reduction, and screening out a subset training decision tree with lower dimensionality through the obtained 3 feature areas;
s23-4: random sample balance dataset D new A subset of samples was obtained, the non-extracted dataset was noted as OT dataset, and the probability of each sample not being extracted was about 0.368. The ratio of each decision tree to the correct classification of the OT data is obtained through the OT data set, the weight value of each decision tree is calculated, and the calculation formula is as follows:
wherein F is j For the classification ratio, K represents a positive integer;
and selecting the decision tree with the highest weight as a final classification result to generate an optimized data set.
S24: dividing the obtained optimized data set into a training set and a testing set (the training set accounts for 70% of the optimized data set, the testing set accounts for 30% of the optimized data set), establishing an improved random forest algorithm model, using a square error as a model prediction precision standard, determining model parameters by using a contrast divergence training algorithm, normalizing the testing set by using a mean_train and a variance std_train obtained by normalizing the training set, wherein a normalization formula of the testing set is as follows:
where x_test is the test set data.
The prediction precision of the model is checked by using a test set, as shown in fig. 2, if the error between the predicted pressure value and the actual measured pressure value of the model reaches the prediction precision requirement (0.97), the current model is judged to be the improved random forest algorithm model.
The method for establishing the fitting pressure function model comprises the following steps: establishing a monitoring point pressure, concentration and flow function relation by adopting a data fitting mode: p=f (Q, C V ) Fitting the functional relationship of the various structures as follows:
P 2 ＝a+bx 1 +clnx 2
P 3 ＝a+bx 1 +cx 2
P 4 ＝a+blnx 1 +clnx 2
as shown in Table 1, the standard error and the goodness of fit of various functional relationships are compared, and the fit pressure function is selected according to the principles of small standard error and large goodness of fit, so that the standard error is ensured to be within 0.006. Function P 1 The minimum standard error is 0.0051 and the maximum fitting goodness is the same, P is selected 1 As a function of fitting pressure, as shown in table 2 and fig. 3.
Table 1 four fitting function evaluation indexes
| Rank | Model | StdError | Residual Sum | Residual Avg. | RSS | R^2 | Ra^2 |
| 1 | P 1 | 0.00514 | 4.56302E-14 | 7.60503E-16 | 0.001479 | 0.94285 | 0.93978 |
| 2 | P 2 | 0.00524 | -2.9865E-14 | -4.9775E-16 | 0.001551 | 0.94007 | 0.93796 |
| 3 | P 3 | 0.00522 | -1.3223E-13 | -2.20379E-15 | 0.001553 | 0.94002 | 0.93792 |
| 4 | P 4 | 0.00522 | -4.6685E-13 | -7.78081E-15 | 0.001555 | 0.93994 | 0.93783 |
TABLE 2 function P 1 Confidence interval
After the fitting pressure function is selected, data fitting is carried out on different key monitoring points respectively:
wherein PT A1 For the pressure predicted value at the monitoring point corresponding to the minimum line pressure, a=108.53, b= -3.094, c=2.188, d=7.822, x 1 To fill concentration data in a station, x 2 Is the data of the outlet flow of the stirring barrel.
Wherein PT B1 For the predicted value of the pressure at the monitoring point corresponding to the maximum value of the pipeline pressure, a '=109.41, b' = -2.04, c '=2.142, d' = 8.236, x1 is the concentration data in the filling station, and x2 is the outlet flow data of the stirring barrel.
S3: and judging the prediction precision of the improved random forest algorithm model and the fitting pressure function model, if the prediction precision meets the standard, executing the step S4, otherwise, returning to the step S1, and re-acquiring data to build the model.
Preferably, the method for judging and improving the prediction precision of the random forest algorithm model and the fitting pressure function model comprises the following steps: and (3) carrying out mean square error calculation on the pressure predicted value of the improved random forest algorithm model and the pressure predicted value of the fitted pressure function model based on the pressure predicted value of the improved random forest algorithm model, and judging that the prediction precision meets the standard if the calculated value is within a set range (such as 5%). The calculation formula is as follows:
wherein F is A (P) is a pressure predictive value for improving a random forest algorithm model, F A1 And (P) is a pressure predicted value fitting to the pressure function model.
S4: and constructing a pressure prediction model, predicting the pressure of the key monitoring point and carrying out pipeline fault early warning.
Preferably, the pressure prediction model is: the pressure predicted value of the improved random forest algorithm model and the pressure predicted value of the fitted pressure function model respectively occupy a certain weight and are used as the pressure predicted value when the slurry reaches the pressure monitoring point, and the final pressure predicted value of the monitoring point is obtained. For example, taking the average value of the pressure predicted values of the two models as a final pressure predicted value, and if the final pressure predicted value is not within a set normal pressure range, performing fault early warning, and adjusting early warning for the pressure control of the subsequent filling pipeline in advance.
The method is based on intelligent algorithm comprehensive data fitting, a high-precision filling pipeline fault early warning method is formed through a large amount of industrial data, the pipeline pressure state can be dynamically decided and judged, the pipeline pressure state is predicted in advance in the filling slurry conveying process, and a reference is provided for follow-up slurry conveying control in advance.
It is to be understood that the above examples of the present invention are provided for clarity of illustration only and are not limiting of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (8)
1. The high-precision filling pipeline fault early warning method based on the artificial intelligence algorithm is characterized by comprising the following steps of:
s1: collecting the concentration and flow data of a surface filling station, and obtaining the pressure and time data of key monitoring points of underground pipelines;
s2: establishing an improved random forest algorithm model and a fitting pressure function model;
the method for establishing the improved random forest algorithm model in the step S2 comprises the following steps:
s21: normalizing the concentration, flow, monitoring point pressure and time data obtained in the step S1 to obtain an original data set;
s22: preprocessing an original data set in a mode of combining an improved SMOTE algorithm with mixed sampling to obtain a balanced data set;
s23: performing feature weighting and dimension reduction on the balanced data set through a ReLieff algorithm, screening out a subset training decision tree with a lower effective dimension, and obtaining an optimized data set by adopting a weighted voting method;
s24: dividing an optimized data set into a training set and a testing set, establishing an improved random forest algorithm model, using a square error as a model prediction precision standard, determining model parameters by using a contrast divergence training algorithm, normalizing the testing set by using a mean_train and a variance std_train obtained by normalizing the training set, checking model prediction precision by using the testing set, and judging the current model as the improved random forest algorithm model if the error between a model prediction pressure value and an actual measurement pressure value meets a prediction precision requirement;
s3: judging the prediction precision of the improved random forest algorithm model and the fitting pressure function model, if the prediction precision meets the standard, executing the step S4, otherwise, returning to the step S1;
s4: and constructing a pressure prediction model, predicting the pressure of the key monitoring point and carrying out pipeline fault early warning.
2. The method for early warning of a fault in a high-precision filling pipeline based on an artificial intelligence algorithm as set forth in claim 1, wherein the step S22 specifically includes:
s22-1: clustering the minority class by using a clustering algorithm, determining the distribution state of each cluster in the minority class, then calculating cluster centers, and interpolating and adding sample data on a cluster center and intra-cluster sample connection line according to the following formula:
X new ＝C i +rand(0,1)*(X-C i )i＝1，2...N
wherein X is new To insert samples newly, C i Is a cluster center, X is C i The method comprises the steps of taking an original sample in cluster core clustering;
s22-2: sequentially calculating the sampling number of most types of samples and few types of samples:
P mix ＝Q-L
PM＝P mix ·t
PN＝P mix ·(1-t)
wherein Q is the number of most samples, L is the number of few samples, P mix PM is undersampled quantity, PN is oversampled quantity, and t is a mixing proportion coefficient;
s22-3: undersampling is carried out on most types of samples, oversampling is carried out on few types of samples, brand new data sets are respectively obtained, and the brand new data sets are combined to obtain a balanced data set D new 。
3. The high-precision filling pipeline fault early warning method based on the artificial intelligence algorithm as claimed in claim 1, wherein the step S23 specifically comprises:
s23-1: will balance data set D new The characteristic weight of each sample is set to 0;
s23-2: randomly selecting a sample R from the balanced data set samples, selecting k similar nearest neighbor samples of the sample R from similar sample sets of the sample R, respectively selecting k different types of nearest neighbor samples from different types of sample sets of each sample R, and calculating to obtain a characteristic weight corresponding to each type in the sample;
s23-3: repeating the step S23-2 for m times to obtain m feature weights of each class, calculating a feature weight average value of each class as a weight value W of each class, arranging the classes in descending order according to the weight values W, uniformly dividing the features into high, medium and low 3 areas according to the weight values, carrying out clustering reduction, and screening out a subset training decision tree with lower dimensionality through the obtained 3 feature areas;
s23-4: random samplingBalancing data set D new Obtaining a sample subset, recording an unextracted data set as an OT data set, obtaining the ratio of each decision tree to the correct classification of the OT data through the OT data set, and calculating the weight value of each decision tree, wherein the calculation formula is as follows:
wherein F is j Is a classification ratio;
and selecting the decision tree with the highest weight as a final classification result to generate an optimized data set.
4. The method for early warning of a fault in a high-precision filling pipeline based on an artificial intelligence algorithm as claimed in claim 1, wherein the method for establishing a fitting pressure function model in step S2 is as follows: establishing a monitoring point pressure, concentration and flow function relation by adopting a data fitting mode: p=f (Q, C V ) Fitting the functional relations of various structures, comparing standard errors and fitting goodness of the functional relations, and selecting a fitting pressure function according to the principles of small standard errors and large fitting goodness; after the fitting pressure function is selected, data fitting is carried out on different monitoring points respectively.
5. The method for early warning of a fault in a high-precision filling pipeline based on an artificial intelligence algorithm according to claim 1, wherein the method for judging and improving the prediction precision of a random forest algorithm model and a fitting pressure function model in the step S3 is as follows: and carrying out mean square error calculation on the pressure predicted value of the improved random forest algorithm model and the pressure predicted value of the fitted pressure function model, and judging that the prediction precision reaches the standard if the calculated value is in a set range.
6. The high-precision filling pipeline fault early warning method based on the artificial intelligence algorithm as claimed in claim 1, wherein the pressure prediction model in the step S4 is as follows: and the pressure predicted value of the improved random forest algorithm model and the pressure predicted value of the fitted pressure function model respectively occupy a certain weight to obtain the final pressure predicted value of the monitoring point.
7. The high-precision filling pipeline fault early warning method based on the artificial intelligence algorithm as claimed in claim 1, wherein the method comprises the following steps of: and the step S1 further comprises preprocessing the acquired data, carrying out data fusion processing by taking the time difference of the ground surface slurry reaching the pressure monitoring point as a reference, and eliminating system abnormality and parking data.
8. The method for early warning of a fault in a high-precision filling pipeline based on an artificial intelligence algorithm according to any one of claims 1 to 7, wherein the method for acquiring pressure and time data of key monitoring points of a downhole pipeline in step S1 is as follows: determining a pressure value generated by potential energy of slurry at a lower port of a pipeline in a vertical section, subtracting a slurry along-path pressure loss value obtained by a loop test, and obtaining a slurry pressure value at any point of the pipeline; and taking the positions corresponding to the maximum pressure value and the minimum pressure value as key point positions of the underground filling pipeline, installing a pressure transmitter at the key point positions to measure pressure data, and recording the pressure and time data by using a paperless recorder.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210683749.5A CN114912372B (en) | 2022-06-17 | 2022-06-17 | High-precision filling pipeline fault early warning method based on artificial intelligence algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210683749.5A CN114912372B (en) | 2022-06-17 | 2022-06-17 | High-precision filling pipeline fault early warning method based on artificial intelligence algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114912372A CN114912372A (en) | 2022-08-16 |
| CN114912372B true CN114912372B (en) | 2024-01-26 |
Family
ID=82771201
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210683749.5A Active CN114912372B (en) | 2022-06-17 | 2022-06-17 | High-precision filling pipeline fault early warning method based on artificial intelligence algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114912372B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115879049A (en) * | 2023-03-02 | 2023-03-31 | 国网江西省电力有限公司电力科学研究院 | Induction identification output method and system of automatic vertical rod |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110107815A (en) * | 2018-11-08 | 2019-08-09 | 清华大学合肥公共安全研究院 | The leakage detection method and device of gas pipeline |
| CN110991666A (en) * | 2019-11-25 | 2020-04-10 | 远景智能国际私人投资有限公司 | Fault detection method, model training method, device, equipment and storage medium |
| CN111985571A (en) * | 2020-08-26 | 2020-11-24 | 国网湖南省电力有限公司 | Low-voltage intelligent monitoring terminal fault prediction method, device, medium and equipment based on improved random forest algorithm |
| CN112819097A (en) * | 2021-02-26 | 2021-05-18 | 浙大城市学院 | Risk evaluation method for hydrogen energy equipment in hydrogen refueling station |
| CN113344075A (en) * | 2021-06-02 | 2021-09-03 | 湖南湖大金科科技发展有限公司 | High-dimensional unbalanced data classification method based on feature learning and ensemble learning |
| CN113483805A (en) * | 2021-06-10 | 2021-10-08 | 中交疏浚技术装备国家工程研究中心有限公司 | Data processing method for long-distance pipeline transportation field dredging transportation system |
-
2022
- 2022-06-17 CN CN202210683749.5A patent/CN114912372B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110107815A (en) * | 2018-11-08 | 2019-08-09 | 清华大学合肥公共安全研究院 | The leakage detection method and device of gas pipeline |
| CN110991666A (en) * | 2019-11-25 | 2020-04-10 | 远景智能国际私人投资有限公司 | Fault detection method, model training method, device, equipment and storage medium |
| WO2021107866A1 (en) * | 2019-11-25 | 2021-06-03 | Envision Digital International Pte. Ltd. | Method and apparatus for detecting fault, method and apparatus for training model, and device and storage medium |
| CN111985571A (en) * | 2020-08-26 | 2020-11-24 | 国网湖南省电力有限公司 | Low-voltage intelligent monitoring terminal fault prediction method, device, medium and equipment based on improved random forest algorithm |
| CN112819097A (en) * | 2021-02-26 | 2021-05-18 | 浙大城市学院 | Risk evaluation method for hydrogen energy equipment in hydrogen refueling station |
| CN113344075A (en) * | 2021-06-02 | 2021-09-03 | 湖南湖大金科科技发展有限公司 | High-dimensional unbalanced data classification method based on feature learning and ensemble learning |
| CN113483805A (en) * | 2021-06-10 | 2021-10-08 | 中交疏浚技术装备国家工程研究中心有限公司 | Data processing method for long-distance pipeline transportation field dredging transportation system |
Non-Patent Citations (1)
| Title |
|---|
| Random Forest Slurry Pressure Loss Model Based on Loop Experiment;Zengjia Wang et al.;Minerals;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114912372A (en) | 2022-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111046341A (en) | Unconventional natural gas fracturing effect evaluation and capacity prediction method based on principal component analysis | |
| CN108053148B (en) | Efficient fault diagnosis method for power information system | |
| CN112364560B (en) | Intelligent prediction method for working hours of mine rock drilling equipment | |
| CN107403279B (en) | Oil transfer pump working condition self-adaptive state early warning system and method | |
| CN114912372B (en) | High-precision filling pipeline fault early warning method based on artificial intelligence algorithm | |
| CN111738348B (en) | Power data anomaly detection method and device | |
| CN104462808A (en) | Method for fitting safe horizontal displacement and dynamic data of variable sliding window of water level | |
| CN111639882A (en) | Deep learning-based power utilization risk judgment method | |
| CN111680407B (en) | Satellite health assessment method based on Gaussian mixture model | |
| CN114266289A (en) | Complex equipment health state assessment method | |
| CN110874685A (en) | Intelligent electric energy meter running state distinguishing method and system based on neural network | |
| CN113484817A (en) | Intelligent electric energy meter automatic verification system abnormity detection method based on TSVM model | |
| CN111369056B (en) | Geological disaster prediction method and electronic equipment | |
| CN115130578A (en) | Incremental rough clustering-based online evaluation method for state of power distribution equipment | |
| CN114021620B (en) | BP neural network feature extraction-based electric submersible pump fault diagnosis method | |
| CN109447026B (en) | Automatic pile body integrity category identification method based on support vector machine | |
| CN105139287A (en) | Comprehensive electric energy metering device state assessment method integrating safety domain | |
| CN116957534A (en) | Method for predicting replacement number of intelligent electric meter | |
| CN116629686A (en) | Method and device for evaluating enterprise energy consumption data | |
| CN113687411B (en) | Earth stress azimuth prediction method based on microseism | |
| Zhai et al. | Parameter estimation method of mixture distribution for construction machinery | |
| CN115081485A (en) | AI-based automatic analysis method for magnetic flux leakage internal detection data | |
| CN114818221A (en) | Pipe network operation situation analysis method based on pipe network hydrodynamic model and data driving model | |
| CN113191556A (en) | Nuclear power Loca event fault prediction and diagnosis method | |
| CN112445632A (en) | HPC reliability evaluation method based on fault data modeling |
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 |