CN105203150B - A kind of chemical plant installations instrumented data exception point-type lapse error detection method - Google Patents
A kind of chemical plant installations instrumented data exception point-type lapse error detection method Download PDFInfo
- Publication number
- CN105203150B CN105203150B CN201510579123.XA CN201510579123A CN105203150B CN 105203150 B CN105203150 B CN 105203150B CN 201510579123 A CN201510579123 A CN 201510579123A CN 105203150 B CN105203150 B CN 105203150B
- Authority
- CN
- China
- Prior art keywords
- data
- point
- wavelet
- error detection
- time
- 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
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention provides a kind of chemical plant installations instrumented data exception point-type lapse error detection method, wavelet method is combined by the detection method with box figure method, chemical industry process instrument data are as sample data in selection a period of time, wavelet transformation is carried out to sample data first, local characteristicses according to sample data time-frequency domain, obtain wavelet conversion coefficient of the data under each yardstick, then rejecting outliers are carried out to the wavelet conversion coefficient under each yardstick by box figure method, pass through the relation mapped one by one, determine the position of the abnormal point-type human error point of original meter data, realize chemical process instrumented data exception point-type lapse error detection.Detection method provided by the invention, the time-frequency domain local characteristicses of application data, overcome erroneous judgement of the prior art to abnormity point, applicability is stronger, overcomes the dependence to data normal distribution in existing method.
Description
Technical field
The present invention relates to the data processing method of chemical plant installations instrument, more particularly to a kind of chemical plant installations instrumented data are abnormal
Point-type lapse error detection method.
Background technology
Accurate measurement data is to implement the basis of control and optimization operation to chemical process.In order to grasp chemical process
Operation characteristic, improve production operation, often to analyze the measurement data from some unit even complex process device.
But generally, unavoidably there are various errors in the measurement data of chemical process so that measurement data can not accurately expire
In sufficient process physics and chemical rule, such as stoichiometric relationship, mass balance and heat balance.Quickly, it is reliable to miss
Poor detection algorithm, significance is suffered from for theory analysis and actual production operation.
According to the difference of measurement error producing method, random error and the major class of human error two can be divided into.The former by with
The influence of machine factor and produce, obey certain statistical law, be inevitable.The latter is probably then to be lost by measuring instrumentss
Caused by the failures such as spirit, fluctuation of service, equipment leakage.In general, the human error of dynamical system is divided into two classes：System is inclined
Poor type human error and abnormal point-type human error.System deviation type human error refers to, within a period of time, measured value is always
The amplitude of some fixation more high or low than actual value, it may be possible to as caused by the factors such as instrument drift, process leaks；It is abnormal
Point-type human error refers to, there is the point of other data of substantial deviation in measurement data, such as spike in measurement signal etc., may
It is to be caused by factors such as instrument measurements.
Instrumented data Normal Distribution is assumed mostly to the detection method of instrumented data exception point-type human error at present,
Using 3 σ methods, i.e. data definition outside the σ of data distribution average positive and negative deviation 3 is exceptional value.But actual instrument data one
As do not meet normal distribution, so method detection abnormal point-type human error point accuracy it is poor.Another box figure method
(boxplot) independent of instrumented data Normal Distribution it is assumed that carrying out abnormity point by the quartile of data distribution
Judge, versatility is stronger.Yet with the dependence not accounted between time upper adjacent data, the detection to abnormity point is easy
Judge by accident.Wavelet method, that is, according to the change of instrumented data frequency spectrum, judges instrument by the way that data are transformed into frequency domain from time-domain
The local change of table data.
The content of the invention
The problem of for existing detection method accuracy difference to instrumented data exception point-type human error, the present invention will
Wavelet method is combined with box figure method, it is proposed that a kind of chemical plant installations instrumented data exception point-type lapse error detection method.
The present invention uses following technical scheme：
A kind of chemical plant installations instrumented data exception point-type lapse error detection method, including：
Step 1：Sample data is collected, a server is configured in the control room of chemical enterprise, server fills with production
The live database server put is connected, and live database server gathers the data of the instrument from production scene, Mei Geyi
The individual sampling period records a real time data, continuously collects the data in multiple sampling periods, then forms sample data, the sample
Input of the data as abnormal point-type lapse error detection method；
Step 2：Choose suitable wavelet transform function, including Haar small echos, dbN wavelet systems, symN wavelet systems, Morlet
It is small to involve Meyer small echos；
Step 3：Sample different frequency scale resolution and wavelet transformation is carried out to sample data, obtain small under different scale
Wave conversion coefficient, the different local circumstances in reflected sample data time domain；
Step 4：Wavelet conversion coefficient under each yardstick is analyzed, radio-frequency component corresponds to be become soon in time domain
Point；
Step 5：To the wavelet conversion coefficient application box figure method of each yardstick obtained under wavelet transformation, wavelet conversion coefficient
As input data, five statistics of input data, i.e. data minimum value, maximum, median, first quartile are calculated
Q1And the 3rd quartile Q2, wherein first quartile Q1It is the median of the data between minimum value and median, the three or four
Quantile Q2It is the median of data between maximum and median；
Step 6：Calculate the spacing I of quartileQR=Q3-Q1, the bound of the wavelet conversion coefficient of each yardstick is calculated,
Wherein the upper limit is defined as HIGH=Q3+m*IQR, lower limit is defined as LOW=Q1-n*IQR, wherein, m and n are fixed coefficient；
Step 7：When data x ＞ HIGH or x ＜ LOW, x are abnormity point, and at the time of record current x values and correspond to；
Step 8：At the time of the abnormity point detected according to box figure method corresponds to, pass through data point in wavelet transformation time-frequency domain
Mapping relations one by one, the position of the abnormal point-type human error point of original meter data is determined, realizes abnormal point-type human error
Detection.
The invention has the advantages that：
A kind of chemical plant installations instrumented data exception point-type lapse error detection method of the present invention, by wavelet method and box
Figure method is combined, and chooses chemical industry process instrument data in a period of time, carries out wavelet transformation to data first, according to data time-frequency
The local characteristicses in domain, wavelet conversion coefficient of the data under each yardstick is obtained, then by box figure method to small under each yardstick
Wave conversion coefficient carries out rejecting outliers, by the relation mapped one by one, determines that the abnormal point-type fault of original meter data misses
Not good enough position, realize chemical process instrumented data exception point-type lapse error detection, the time-frequency domain office of application data of the present invention
Portion's characteristic, overcome erroneous judgement of the prior art to abnormity point, applicability is stronger, overcomes in existing method to data normal distribution
Dependence.
Embodiment
The present invention is specifically described below：
A kind of chemical plant installations instrumented data exception point-type lapse error detection method, including：
Step 1：Sample data is collected, a server is configured in the control room of chemical enterprise, server fills with production
The live database server put is connected, and live database server gathers the data of the instrument from production scene, Mei Geyi
The individual sampling period records a real time data, continuously collects the data in multiple sampling periods, then forms sample data, the sample
Input of the data as abnormal point-type lapse error detection method；
Step 2：Choose suitable wavelet transform function, including Haar small echos, dbN wavelet systems, symN wavelet systems, Morlet
It is small to involve Meyer small echos etc.；
Step 3：Sample different frequency scale resolution and wavelet transformation is carried out to sample data, obtain small under different scale
Wave conversion coefficient, the different local circumstances in reflected sample data time domain, typically uses level Four change of scale；
Step 4：Wavelet conversion coefficient under each yardstick is analyzed, radio-frequency component often corresponds to the fast change in time domain
Composition, such as steep front, rear edge and spike, as need the abnormal point-type human error detected；
Step 5：To the wavelet conversion coefficient application box figure method of each yardstick obtained under wavelet transformation, wavelet conversion coefficient
As input data, five statistics of input data, i.e. data minimum value, maximum, median, first quartile are calculated
Q1And the 3rd quartile Q2, wherein first quartile Q1It is the median of the data between minimum value and median, the three or four
Quantile Q2It is the median of data between maximum and median；
Step 6：Calculate the spacing I of quartileQR=Q3-Q1, the bound of the wavelet conversion coefficient of each yardstick is calculated,
Wherein the upper limit is defined as HIGH=Q3+m*IQR, lower limit is defined as LOW=Q1-n*IQR, wherein, m and n are fixed coefficient；
Step 7：When data x ＞ HIGH or x ＜ LOW, x are abnormity point, and at the time of record current x values and correspond to；
Step 8：At the time of the abnormity point detected according to box figure method corresponds to, pass through data point in wavelet transformation time-frequency domain
Mapping relations one by one, the position of the abnormal point-type human error point of original meter data is determined, realizes abnormal point-type human error
Detection.
The chemical plant installations instrumented data exception point-type lapse error detection method, wavelet method is combined with box figure method, real
Existing chemical process instrumented data exception point-type lapse error detection.The time-frequency domain local characteristicses of application data of the present invention, overcome existing
There is erroneous judgement of the technology to abnormity point, applicability is stronger, overcomes the dependence to data normal distribution in existing method.
Certainly, described above is not limitation of the present invention, and the present invention is also not limited to the example above, this technology neck
The variations, modifications, additions or substitutions that the technical staff in domain is made in the essential scope of the present invention, it should also belong to the present invention's
Protection domain.
Claims (1)
- A kind of 1. chemical plant installations instrumented data exception point-type lapse error detection method, it is characterised in that including：Step 1：Collection sample data, one server of configuration in the control room of chemical enterprise, server and process units Live database server is connected, and live database server gathers the data of the instrument from production scene, is adopted every one Real time data of sample periodic recording, continuously collects the data in multiple sampling periods, then forms sample data, the sample data Input as abnormal point-type lapse error detection method；Step 2：Choose suitable wavelet transform function, including Haar small echos, dbN wavelet systems, symN wavelet systems, Morlet small echos And Meyer small echos；Step 3：Sample different frequency scale resolution and wavelet transformation is carried out to sample data, the small echo obtained under different scale becomes Change coefficient, the different local circumstances in reflected sample data time domain；Step 4：Wavelet conversion coefficient under each yardstick is analyzed, radio-frequency component corresponds to the fast change composition in time domain；Step 5：To the wavelet conversion coefficient application box figure method of each yardstick obtained under wavelet transformation, wavelet conversion coefficient conduct Input data, calculate five statistics of input data, i.e. data minimum value, maximum, median, first quartile Q1And 3rd quartile Q2, wherein first quartile Q1It is the median of the data between minimum value and median, the 3rd quartile Number Q2It is the median of data between maximum and median；Step 6：Calculate the spacing I of quartileQR=Q3-Q1, the bound of the wavelet conversion coefficient of each yardstick is calculated, wherein The upper limit is defined as HIGH=Q3+m*IQR, lower limit is defined as LOW=Q1-n*IQR, wherein, m and n are fixed coefficient；Step 7：When data x ＞ HIGH or x ＜ LOW, x are abnormity point, and at the time of record current x values and correspond to；Step 8：At the time of the abnormity point detected according to box figure method corresponds to, by data point in wavelet transformation time-frequency domain one by one Mapping relations, the position of the abnormal point-type human error point of original meter data is determined, realizes abnormal point-type lapse error detection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510579123.XA CN105203150B (en) | 2015-09-11 | 2015-09-11 | A kind of chemical plant installations instrumented data exception point-type lapse error detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510579123.XA CN105203150B (en) | 2015-09-11 | 2015-09-11 | A kind of chemical plant installations instrumented data exception point-type lapse error detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105203150A CN105203150A (en) | 2015-12-30 |
| CN105203150B true CN105203150B (en) | 2017-11-10 |
Family
ID=54950957
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510579123.XA Active CN105203150B (en) | 2015-09-11 | 2015-09-11 | A kind of chemical plant installations instrumented data exception point-type lapse error detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105203150B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108763346B (en) * | 2018-05-15 | 2022-02-01 | 中南大学 | Abnormal point processing method for sliding window box type graph median filtering |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102288286B (en) * | 2011-06-16 | 2012-10-03 | 中国科学院沈阳自动化研究所 | Method for analyzing and evaluating measure point precision of gearbox in vibration acceleration sensor |
| CN102488516A (en) * | 2011-12-13 | 2012-06-13 | 湖州康普医疗器械科技有限公司 | Nonlinear electroencephalogram signal analysis method and device |
| US9122786B2 (en) * | 2012-09-14 | 2015-09-01 | Software Ag | Systems and/or methods for statistical online analysis of large and potentially heterogeneous data sets |
| CN104316317A (en) * | 2014-10-08 | 2015-01-28 | 西北工业大学 | Gear system multi-fault diagnosis method based on COM assemblies |
-
2015
- 2015-09-11 CN CN201510579123.XA patent/CN105203150B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN105203150A (en) | 2015-12-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109558295A (en) | A kind of performance indicator method for detecting abnormality and device | |
| CN109284933B (en) | Electronic transformer state evaluation system and method based on mathematical statistics | |
| CN101533052B (en) | Testing system and method for PWM fan electrical performance | |
| CN105116369A (en) | Radio frequency radiation disturbance rejection degree test apparatus and system | |
| CN107765206B (en) | Method for evaluating running state of power quality monitoring device | |
| CN106771582A (en) | The method of testing and tester of high-frequency pulse signal | |
| CN105068035B (en) | A kind of voltage transformer error horizontal dynamic detection method and system | |
| CN104569886B (en) | Signal detection apparatus calibration method based on time and frequency parameter reference instrument mode | |
| CN105203150B (en) | A kind of chemical plant installations instrumented data exception point-type lapse error detection method | |
| CN105547632B (en) | A method of impact velocity is examined and determine or calibrated using impact velocity measuring device | |
| CN103743878A (en) | Water quality monitoring data processing method and device | |
| CN109359822A (en) | Electronic type voltage transformer measuring state appraisal procedure and system | |
| CN108801320B (en) | Diagnosis method of natural gas measurement system | |
| KR20140109596A (en) | Apparatus method for predicting and correcting meter reading data in a range not reading | |
| KR101886210B1 (en) | Measuring instrument reliability evaluation apparatus and operating method thereof | |
| CN105842566B (en) | The system and its mapping method of automation mapping digital filter amplitude-versus-frequency curve | |
| CN111562037A (en) | Thermometer fault detection method and device | |
| CN108957174B (en) | Voltage sag detection device and method | |
| CN108195339B (en) | Uncertainty determination method for plate flatness measurement result in field environment | |
| CN108536777A (en) | A kind of data processing method, server cluster and data processing equipment | |
| CN105445553A (en) | Method for precisely acquiring short-circuit impedance value of transformer at 50Hz | |
| CN102354295A (en) | CPU (central processing unit) load testing method for equipment in DCS (distributed control system) | |
| Ahn et al. | Measurement system analysis for semiconductor measurement process | |
| CN106096219B (en) | A kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance | |
| CN117872474B (en) | Seismic early warning station network waveform data quality analysis method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Wang Chunli Inventor after: Li Chuankun Inventor after: Gao Xinjiang Inventor before: Zhu Jianfeng Inventor before: Wang Chunli Inventor before: Li Chuankun Inventor before: Gao Xinjiang |
|
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20170927 Address after: Yanan City, Shandong province Qingdao City three road 266071 No. 218 Applicant after: Qingdao Safety Engineering Research Institute of Sinopec Co., Ltd. Applicant after: Sinopec Corp. Address before: 100728 Chaoyangmen street, Beijing, No. 22, No. Applicant before: Sinopec Corp. Applicant before: Qingdao Safety Engineering Research Institute of Sinopec Co., Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |