CN112381235A - Auxiliary method for monitoring state change of preheater in cement plant - Google Patents

Auxiliary method for monitoring state change of preheater in cement plant Download PDF

Info

Publication number
CN112381235A
CN112381235A CN202011264149.2A CN202011264149A CN112381235A CN 112381235 A CN112381235 A CN 112381235A CN 202011264149 A CN202011264149 A CN 202011264149A CN 112381235 A CN112381235 A CN 112381235A
Authority
CN
China
Prior art keywords
preheater
machine learning
selecting
cement plant
monitoring
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.)
Withdrawn
Application number
CN202011264149.2A
Other languages
Chinese (zh)
Inventor
樊融
陈向阳
宋志刚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huanqing Energy Technology Shanghai Co ltd
Original Assignee
Huanqing Energy Technology Shanghai Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huanqing Energy Technology Shanghai Co ltd filed Critical Huanqing Energy Technology Shanghai Co ltd
Priority to CN202011264149.2A priority Critical patent/CN112381235A/en
Publication of CN112381235A publication Critical patent/CN112381235A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00—Measuring or testing not otherwise provided for
    • G01D21/02—Measuring two or more variables by means not covered by a single other subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention provides an auxiliary method for monitoring the state change of a preheater of a cement plant, which comprises the following steps: selecting machine learned features and predicted targets: selecting the parameters of the operation of the preheater as the characteristics of machine learning, and selecting the future value of the pressure difference in the preheater as a prediction target; selecting a machine learning method: after the characteristics and the prediction target of machine learning are selected, selecting multi-characteristic time sequence input and providing a machine learning method of the prediction target time sequence; model training: after the machine learning method is selected, model training is carried out; selecting an anomaly criterion: after the model training is finished, establishing an abnormal judgment criterion; deployment and operation: and the model is deployed in the isolation environment, and calculation is performed after data is acquired, so that the normal operation of the facility is not influenced. By adopting the method, an auxiliary means for monitoring the working state change of the preheater in the cement plant is established, and the judgment of the state of the preheater by an operator is facilitated.

Description

Auxiliary method for monitoring state change of preheater in cement plant
Technical Field
The invention relates to the technical field of preheater monitoring, in particular to an auxiliary method for monitoring the state change of a preheater in a cement plant.
Background
At present, cement plant preheaters usually adopt cyclone preheaters, while a cyclone preheater system mainly comprises a rotary kiln and a cyclone behind the kiln, the working process of the system is completed by the cooperation of a plurality of stages of preheaters, and the preheating mode is that raw materials are heated by blowing. However, in the operation process of the preheater, the traditional method adopts manual monitoring to set and manually monitor the temperature, the humidity, the feeding amount and other parameters of the preheater, and is not applied to an auxiliary monitoring automatic control method for the state of the preheater in a cement plant.
Through retrieval, patent document CN109631075A discloses an anti-blocking ash air volume adjusting method and an operation monitoring device for an air preheater, and the prior art discloses an anti-blocking ash air volume adjusting method, which provides a direct basis for adjusting the anti-blocking ash air volume of the air preheater, determines the most appropriate air volume, improves the economical efficiency of the operation of an anti-blocking ash system of the air preheater, and ensures the anti-blocking ash effect. However, the technology described in the patent does not adopt a machine learning method, and needs to install additional monitoring equipment, so that the cost is high.
Through search, patent document CN110779745A discloses a heat exchanger early fault diagnosis method based on a BP neural network, which includes the steps of: A) collecting key parameters of a heat exchanger in a normal working state; B) dividing the acquired key parameters into a training set, a verification set and a test set; C) constructing a BP neural network model; D) obtaining a mean value mu and a standard deviation sigma of a test output error; E) predicting the outlet data of the heat exchanger by using the trained BP neural network model to obtain the predicted output of the outlet data of the heat exchanger; F) and carrying out fault diagnosis. The technique described in this patent is only for heat exchangers, the algorithm is limited to BP neural networks, and the results appear to be non-intuitive to the operator.
Therefore, the auxiliary monitoring method for the state of the preheater in the cement plant needs to be solved urgently, so that an auxiliary means for monitoring the working state change of the preheater in the cement plant can be established, and an operator can conveniently judge the state of the predictor.
Disclosure of Invention
In view of the defects in the prior art, the invention aims to provide an auxiliary method and an auxiliary system for monitoring the state change of a preheater of a cement plant. The model of a certain level of preheater in the cement plant is learned when in a normal state, and the state change of the preheater is judged by combining the operation experience of an operator by comparing the long-term deviation change trend of the predicted value and the actual value of the model, so that the operator is assisted in judging the state degradation condition of the preheater.
According to the auxiliary method for monitoring the state change of the preheater of the cement plant, provided by the invention, the model characteristics of the preheater are learned by utilizing a machine learning method, the trained model is used for predicting the target characteristics, and the state change of the preheater is judged according to the deviation change trend of a predicted value and a true value and by combining the experience of an operator, and the method comprises the following steps:
step 1, selecting characteristics and prediction targets of machine learning: selecting the parameters of the operation of the preheater as the characteristics of machine learning, and selecting the future value of the pressure difference in the preheater as a prediction target;
step 2, selecting a machine learning method: after the characteristics and the prediction target of machine learning are selected, selecting multi-characteristic time sequence input and providing a machine learning method of the prediction target time sequence;
step 3, model training: after the machine learning method is selected, model training is carried out;
and 4, selecting an abnormal criterion: after the model training is finished, establishing an abnormal judgment criterion;
step 5, deployment and operation: and the model is deployed in the isolation environment, and calculation is performed after data is acquired, so that the normal operation of the facility is not influenced.
Preferably, in step 1, historical values of temperature, temperature difference, charge and pressure difference in the preheater are selected as characteristics of machine learning.
Preferably, all the characteristic data in step 1 come from existing sensors and DCS equipment, and no additional hardware is required.
Preferably, step 2 selects a two-layer LSTM machine learning method.
Preferably, a Seq2Seq model with hidden layers containing two layers of LSTM is adopted in step 2, and the length of the input sequence and the length of the output sequence can be selected according to actual requirements and hardware performance.
Preferably, in step 2, the length of the input sequence and the length of the prediction sequence can be selected according to the actual situation of the power plant, including the data acquisition frequency and the system operation stability. The invention adopts the input sequence length of 10 and the output sequence length of 5; the training data sample period is 10s, and resampling is performed according to a 1min period, so that an input sequence of 10 points represents 10 minutes of data.
Preferably, in step 3, according to the selection of the lengths of the input sequence and the output sequence in step 2, the MSE is used as a loss function, the data set is segmented and converted into an input format meeting the LSTM, and the model parameters are adjusted according to the training process.
Preferably, in step 4, the adopted abnormality judgment criteria include alarm frequency and deviation trend, and the judgment and processing should be performed for the case that the input data deviates from the training data input range, and the alarm triggering should be limited to the case that the input data is within the training input data range.
Compared with the prior art, the invention has the following beneficial effects:
1. by adopting the method, an auxiliary means for monitoring the working state change of the preheater in the cement plant is established, and the judgment of the state of the preheater by an operator is facilitated.
2. By adopting the method of the invention, new hardware is not needed to be added, a large amount of data provided by the existing measuring equipment is fully utilized, and the method has the advantages of improving the utilization rate of the existing data and lowering the cost.
3. By adopting the method, the deep research on the mechanism of the preheater is avoided, and the qualitative prediction of the running state of the equipment based on data analysis is conveniently and quickly carried out by manufacturers.
4. According to the method, only the target characteristic historical value sequence can be selected as input, the autoregressive prediction model is obtained, and the limitation that only a neural network learning model can be used is avoided.
5. The prediction-alarm step in the invention does not explicitly count the output mean value and the standard deviation, and directly displays the variation trend of the error by using a graph, so that an operator can intuitively judge by combining with the operation experience, and the use habit of the industrial facility operator is better met.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is an overall flow chart of the present invention;
FIG. 2 is a schematic diagram of the LSTM model structure in the present invention;
FIG. 3 is a diagram of the test effect of the trained model of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
As shown in fig. 1, the auxiliary method for monitoring the state change of the preheater in the cement plant provided by the invention utilizes a machine learning method to learn the model characteristics of the preheater, uses a trained model for predicting the target characteristics, and judges the state change of the preheater according to the deviation change trend of a predicted value and a true value and by combining the experience of an operator, and comprises the following steps:
step 1, selecting characteristics and prediction targets of machine learning: selecting the parameters of the operation of the preheater as the characteristics of machine learning, and selecting the future value of the pressure difference in the preheater as a prediction target; the main parameters related to the operation of the preheater are temperature, temperature difference, pressure difference, air volume, material feeding quantity and the like, and parameters directly or indirectly reflecting the operation state of the preheater are selected to be used as characteristics of machine learning according to the actual conditions of facilities. The invention selects temperature, temperature difference, feeding quantity and pressure difference (historical value) as characteristics; the pressure difference (future value) is selected as the prediction target.
Step 2, selecting a machine learning method: after the characteristics and the prediction target of machine learning are selected, selecting multi-characteristic time sequence input and providing a machine learning method of the prediction target time sequence; machine learning methods that accept multi-feature time series input (feature selection see step 1) and give predicted target time series can be used with the solution described in this invention.
Step 3, model training: after the machine learning method is selected, model training is carried out; the training and testing data of the model in the step 2 are preferably data in normal operation after equipment is renovated, and MSE (mean square error) is used as a loss function.
And 4, selecting an abnormal criterion: after the model training is finished, establishing an abnormal judgment criterion; equipment degradation (e.g., skinning) is a gradual process, and in addition to a proper alarm threshold, a time-based evaluation criterion rather than a single prediction effect should be established to reflect the gradual change process of the deviation between the actual preheater model and the model obtained by machine learning, and the adopted and suggested criteria of the invention include alarm frequency and deviation trend.
Step 5, deployment and operation: and the model is deployed in the isolation environment, and calculation is performed after data is acquired, so that the normal operation of the facility is not influenced.
Further, in step 1, all the feature data come from the existing sensors and DCS devices, and no other devices are added. The feature selection should be performed according to the results of correlation analysis, interviews by professional engineers; when the feeding amount, the temperature and the air quantity are stable, the change of the wind resistance possibly caused by the problems of skinning and the like is reflected as the change of the pressure difference;
in step 2, a Seq2Seq model with hidden layers containing two layers of LSTMs is adopted, the length of an input sequence is about 10 (which can be selected according to actual requirements and hardware performance), and the length of an output sequence is about 5 (which can be selected according to actual requirements and hardware performance); the training data sampling period is 10s, and resampling is performed according to the 1min period, so that the 10-point input sequence represents 10 minutes of data.
In step 3, the data set is segmented according to the length selection of the input and output sequences in the step 2, and is converted into an input format meeting the LSTM; model parameters such as the number of units on each layer, activation functions and the like are adjusted according to the training process; and (4) eliminating abnormal data from the input data, normalizing the input data and recording the range of the characteristics.
In step 4, the alarm threshold value, the alarm frequency limit value and the like are continuously adjusted according to the test and commissioning results to match the conditions of specific facilities; the case where the input data deviates from the training data input range should be judged and handled, and the alarm triggering should be limited to the case where the input data is within the training input data range. After the alarm is triggered, the operator judges the state of the equipment according to the operation experience.
In step 5, the deployment and operation should not affect the existing equipment of the facility, and necessary isolation must be performed, such as using a dedicated computing server.
Furthermore, in step 2, a Seq2Seq model with hidden layers containing two layers of LSTM is adopted, and the length of the input sequence and the length of the output sequence can be selected according to actual requirements and hardware performance.
The double-layer LSTM is realized by utilizing a keras library of Tensorflow, and the main parameters of the double-layer LSTM are as follows:
a first layer:
Units 64
Activation Tanh
a second layer:
Units 32
Activation Tanh
the number of units in each layer can be adjusted according to the needs, and other parameters can be set according to the common practice.
Example 1:
according to the idea and steps of the invention, an example is given as follows:
firstly, according to collected data, selecting an inlet temperature, an inlet and outlet temperature difference, a feeding amount and a pressure difference historical value of a 5 th-stage preheater of a certain cement plant as characteristic values, and predicting a target to be a pressure difference;
secondly, setting the length of an input sequence and an output sequence, and segmenting data into a training set and a test set meeting LSTM input by using a developed software tool;
thirdly, establishing a double-layer LSTM prediction model based on a Tensorflow framework, and repeatedly training;
fourthly, testing the model and preparing for deployment;
the structure of the established LSTM model is shown in FIG. 2, the test effect of the trained model is shown in FIG. 3, and note that: the display data points are fewer, and the deviation between the predicted value and the true value becomes larger when the state of the equipment changes along with the time.
After the operation is carried out for a long time, if the mechanism model of the preheater changes due to some factors (such as skinning, blockage and the like), the predicted value and the true value may have larger deviation, and an operator can judge the state of the preheater according to a deviation trend chart and by combining operation experience.
Those skilled in the art will appreciate that, in addition to implementing the system and its various devices, modules, units provided by the present invention as pure computer readable program code, the system and its various devices, modules, units provided by the present invention can be fully implemented by logically programming method steps in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system and various devices, modules and units thereof provided by the invention can be regarded as a hardware component, and the devices, modules and units included in the system for realizing various functions can also be regarded as structures in the hardware component; means, modules, units for performing the various functions may also be regarded as structures within both software modules and hardware components for performing the method.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.

Claims (9)

1. An auxiliary method for monitoring the state change of a preheater of a cement plant is characterized in that a machine learning method is used for learning the model characteristics of the preheater, a trained model is used for predicting target characteristics, and the state change of the preheater is judged according to the deviation change trend of a predicted value and a true value and by combining with the experience of an operator.
2. An auxiliary method of monitoring a change in the condition of a preheater of a cement plant according to claim 1, comprising the steps of:
step 1, selecting characteristics and prediction targets of machine learning: selecting the parameters of the operation of the preheater as the characteristics of machine learning, and selecting the future value of the pressure difference in the preheater as a prediction target;
step 2, selecting a machine learning method: after the characteristics and the prediction target of machine learning are selected, selecting multi-characteristic time sequence input and providing a machine learning method of the prediction target time sequence;
step 3, model training: after the machine learning method is selected, model training is carried out;
and 4, selecting an abnormal criterion: after the model training is finished, establishing an abnormal judgment criterion;
step 5, deployment and operation: and the model is deployed in the isolation environment, and calculation is performed after data is acquired, so that the normal operation of the facility is not influenced.
3. An auxiliary method for monitoring the state change of a preheater of a cement plant as claimed in claim 2, wherein the historical values of the temperature, the temperature difference, the charge amount and the pressure difference in the preheater are selected as the characteristics of machine learning in step 1.
4. The method as claimed in claim 2, wherein all the characteristic data in step 1 are obtained from existing sensors and DCS equipment without adding extra hardware.
5. The auxiliary method for monitoring the state change of the preheater of the cement plant as recited in claim 2, wherein the step 2 selects a machine learning method of double layer LSTM.
6. The auxiliary method for monitoring the state change of the preheater of the cement plant as claimed in claim 5, wherein the Seq2Seq model with the hidden layer containing two layers of LSTM is adopted in the step 2, and the length of the input sequence and the length of the output sequence can be selected according to actual requirements and hardware performance.
7. The auxiliary method for monitoring the state change of the preheater of the cement plant as claimed in claim 2, wherein in the step 2, the length of the input sequence and the length of the prediction sequence can be selected according to the actual condition of the cement plant, including the data acquisition frequency and the stability degree of the system operation.
8. The auxiliary method for monitoring the state change of the preheater of the cement plant as claimed in claim 2, wherein in the step 3, according to the selection of the length of the input sequence and the output sequence in the step 2, MSE is adopted as a loss function, the data set is subjected to segmentation processing, the data set is converted into an input format meeting LSTM, and the model parameters are adjusted according to the training process.
9. An auxiliary method for monitoring the preheater status of cement plant as recited in claim 2, wherein the anomaly determination criteria used in step 4 include alarm frequency and deviation trend, and the input data should be determined and processed for deviation from the training data input range, and the alarm triggering should be limited to the case where the input data is within the training input data range.
CN202011264149.2A 2020-11-12 2020-11-12 Auxiliary method for monitoring state change of preheater in cement plant Withdrawn CN112381235A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011264149.2A CN112381235A (en) 2020-11-12 2020-11-12 Auxiliary method for monitoring state change of preheater in cement plant

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011264149.2A CN112381235A (en) 2020-11-12 2020-11-12 Auxiliary method for monitoring state change of preheater in cement plant

Publications (1)

Publication Number Publication Date
CN112381235A true CN112381235A (en) 2021-02-19

Family

ID=74583493

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011264149.2A Withdrawn CN112381235A (en) 2020-11-12 2020-11-12 Auxiliary method for monitoring state change of preheater in cement plant

Country Status (1)

Country Link
CN (1) CN112381235A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113219871A (en) * 2021-05-07 2021-08-06 淮阴工学院 Curing room environmental parameter detecting system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113219871A (en) * 2021-05-07 2021-08-06 淮阴工学院 Curing room environmental parameter detecting system

Similar Documents

Publication Publication Date Title
EP3809220B1 (en) Method and system for semi-supervised deep anomaly detection for large-scale industrial monitoring systems based on time-series data utilizing digital twin simulation data
CN108604360B (en) Facility abnormity monitoring method and system
CN111412579B (en) Air conditioning unit fault type diagnosis method and system based on big data
CN110794227B (en) Fault detection method, system, device and storage medium
CN110008565A (en) A kind of industrial process unusual service condition prediction technique based on operating parameter association analysis
CN110443488A (en) The recognition methods of drilling well spill hazard, system and equipment based on convolutional neural networks
JP2005339558A (en) Method for developing unified quality assessment and providing automated fault diagnostic tool for turbine machine systems and the like
KR20050067018A (en) Model making device between the relation in process and quality and method thereof
CN112097365A (en) Air conditioner fault detection and identification method and device based on prediction and classification model
CN111581597A (en) Wind turbine generator gearbox bearing temperature state monitoring method based on self-organizing kernel regression model
CN114282443B (en) Residual service life prediction method based on MLP-LSTM supervised joint model
CN110757510B (en) Method and system for predicting remaining life of robot
CN116050665B (en) Heat supply equipment fault prediction method
CN114297909A (en) Water pump fault diagnosis method and system based on neural network
CN111143934A (en) Structural deformation prediction method based on time convolution network
CN112381235A (en) Auxiliary method for monitoring state change of preheater in cement plant
CN117662445B (en) Nitrogen compressor operation control method and system
CN117273402B (en) Energy-saving management system and method for glass deep processing production line based on Internet of Things technology
KR102494420B1 (en) Machine learning-based clean room management system and method
CN108804796B (en) Annular cooler air leakage rate detection method based on frequency spectrum characteristics
CN116235148A (en) Early warning method and device for downtime of dry pump, electronic equipment, storage medium and program
CN108895839B (en) sintering machine air leakage fault online diagnosis method
CN108709426B (en) Sintering machine air leakage fault online diagnosis method based on frequency spectrum characteristic bilateral detection method
CN115496188A (en) Coal mill fault early warning method based on deep learning convolutional neural network
CN108733859B (en) Early warning method for deformation of support plate in heat treatment process of steam generator

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
WW01 Invention patent application withdrawn after publication

Application publication date: 20210219

WW01 Invention patent application withdrawn after publication