CN101424550A - Instrument meter freezing fault rapid detecting method - Google Patents

Abstract

The invention provides a method for quickly detecting the failure of freezing a meter, comprising the following steps that: (1)a sampling data window time of measured variable A is set, n measured variables are acquired, and each measured variable is subjected to gross error elimination; (2) an equation of linear regression in one unknown is established; (3) the meter failure is judged by using a comparison method, comprising that: (3.1) when the meter to be detected is in the adjusting loop, the measured value of the meter to be detected is set to vary in the same direction or opposite direction to the measured value of other meters; when it is detected that the measured values of the meter to be detected and other meters vary in opposite direction or the same direction, the failure of freezing a meter is judged; (3.2) when the meter to be detected is not in the adjusting loop, the variable quantity within the sampling data window time is calculated according to the linear equation (1); when the variable quantity surpasses the upper limit threshold value, the failure of freezing a meter is judged. The invention provides a method for quickly detecting the failure of freezing a meter, which can effectively detect and alarm after the failure of freezing a meter appears.

Description

Instrument meter freezing fault rapid detecting method

Technical field

The present invention relates to the instrument fault detection range, especially a kind of instrument meter freezing fault detection method based on DCS Distributed Control System (DCS) and the advanced control of APC forecasting techniques.

Background technology

In oil, petrochemical production process, water system, the vapour system overwhelming majority are that the measuring instrument sampling line of these systems is as conductive medium substantially with water outdoor open-air.As vapour system pressure, flow measurement, water system pressure, flow measurement etc.In the north, these measuring instrument sampling lines and transmitter all need to add the tracing thermal-insulating system, so that guarantee that in the winter of severe cold aqueous medium does not freeze in the instrument sampling line, it is normal to make it the pressure conduction, thereby makes instrumentation normal.But tracing thermal-insulating system maintenance work amount is very big, and if any the accident that will freeze stopple coupon and transmitter accidentally, we are referred to as meter freezing fault.After meter freezing fault takes place, can cause indication inaccurate, technological parameter departs from even stops, and also can scrap the bursting by freezing of transmitter bellows.

Freeze table and generally occur in 3-4 o'clock morning in winter, also be the technological operation personnel and patrol and examine the workman the most tired the time, at this moment the easiest breaking down.After freezing stopple coupon, it is inaccurate that indication can take place indicating instrument, can cause that operating personnel judge by accident, maloperation.After the indication of regulating loop instrument meter freezing is inaccurate, regulated variable can takes place depart from normal value, bring big disturbance so that stop to process system.After the indication of interlock circuit instrument meter freezing was inaccurate, interlocking action parking accident can take place.The discovery of this class accident is is at present still patrolled and examined by the workman scene and is found and the technological operation personnel find that technological process has problem and recording curve unusual etc., and much this class accident parking has taken place is just known.

In the coldest winter below-20 ℃, after freezing table and taking place, general interlock circuit instrument is stopped the big time about about 30 minutes from beginning to respond to.After meter freezing fault occurred in addition, gradual skew took place in the measured value of instrument, and this is because the transmitter film closes the expansion strength institute of being iced extremely.

The research institution that has at present is engaged in the research of the on-line measurement of instrument fault, in " chemical process automation and instrument " 1 phase in 2008 " process industry instrument fault method for quick " literary composition, adopt DCS opc server universal data interface, data acquisition is carried out computing to host computer, size according to the variance of instrument signal variable quantity is judged instrument fault, leaks faults such as the measurement that causes is inaccurate at the fault of instrument own and analytical table stopple coupon and all has good detection and judge; But, after meter freezing fault occurs, gradual skew takes place in the measured value of instrument, the variance of instrument signal variable quantity is still very big, can't judge this class fault, further, be subjected to the restriction of DCSOPC server data transmission speed, the prior art sampling interval is to the number of minutes magnitude, and obviously for the system of freezing table and stopping soon, meter freezing fault detects difficult accomplishing.

Summary of the invention

In order to overcome the effectively deficiency of measuring instrument meter freezing fault of existing instrument fault detection method, the invention provides a kind of after meter freezing fault occurs the effective instrument meter freezing fault rapid detecting method of detection alarm.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of instrument meter freezing fault rapid detecting method, 1), the sampling data window time of measurand A is set described instrument meter freezing fault rapid detecting method comprises:, gather n measurand, and each measurand is carried out elimination of rough difference;

2), set up the simple regression linear equation, host computer carries out fitting a straight line to the DCS The data one-variable linear regression method in the sampling data window time, the equation of line of match is (1):

$\hat{A}=\hat{a}+\hat{b}t---\left(1\right)$

Wherein,

$\hat{b}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{t}_i{A}_i-n\stackrel{\‾}{t}\stackrel{\‾}{A}}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{t_i}^2-n{\stackrel{\‾}{t}}^2}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(t_i-\stackrel{\‾}{t})(A_i-\stackrel{\‾}{A})}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(t_i-\stackrel{\‾}{t})}^2}$

$\stackrel{\‾}{t}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{t}_i$ $\stackrel{\‾}{A}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{A}_i$

$\hat{a}=\stackrel{\‾}{A}-\hat{b}\stackrel{\‾}{t}$

Slope

The change direction of expression instrument measurement value trend;

3), adopt pairing comparision to judge instrument fault, comprising:

(3.1), when instrument to be detected at regulating loop, set the measured value of described instrument to be detected and the equidirectional or reverse direction variation of measured value of other instrument, when reverse direction or equidirectional variation appear in the measured value that detects instrument to be detected and other instrument, be judged to be meter freezing fault;

(3.2), when instrument to be detected not at regulating loop, according to the variable quantity in straight-line equation (1) the calculating sampling data window time,, be judged to be meter freezing fault when described variable quantity surpasses upper limit threshold.

As preferred a kind of scheme: in step 2) in calculate residual sum of squares (RSS):

$Q=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(A_i-\widehat{A_i})}^2;$

Described fault detection method also comprises:

4), the historical data the when residual sum of squares (RSS) of institute's fitting a straight line and instrument to be detected are normally moved does the ratio computing, less than threshold value, is judged to be meter freezing fault as ratio.

Further, in described step 1), the process of each measurand being carried out elimination of rough difference is:

(1.1), to the measurand calculating mean value in the Preset Time section, establish A _iBe i instrument sampled measurement variable, mean value calculation formula (2) is:

$\stackrel{\‾}{A}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{A}_i---\left(2\right)$

Wherein, A: be mean value, n measures total degree in the Preset Time section;

(1.2), represent the error that calculates after the true value, note is made V with arithmetic mean X mean value _i, V _i=A _i-A

Can get the estimated value of standard deviation δ according to the Bezier formula, calculating formula is (3):

$\mathrm{\δ}\≈\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{v}_i^2}{(n-1)}}---\left(3\right)$

(1.3), if certain remainder error | V _i| 3 σ, judge in this measured value and contain gross error, should give rejecting, with A _iAfter the rejecting, after recomputating mean value and standard deviation, the remaining measured value of usefulness judges by above-mentioned steps that still until there not being gross error, elimination of rough difference finishes.

As preferred another scheme: in described step (1), a secondary data was gathered at the every interval of supercomputing module among the data acquisition system (DAS) DCS in one second, and carrying out pre-service with interior data to 1 minute, the mean value that promptly calculates 10 6 seconds measured values calls for host computer.

Further, when outdoor temperature drops to subzero 2 degrees centigrade, drop into described supercomputing module.

Further, described sampling data window time is along with the outdoor temperature rule change: outdoor temperature is low more, and sampling data window time is short more.

Technical conceive of the present invention is: after meter freezing fault occurred, gradual skew took place in the measured value of instrument, and most applications is that the recording curve of the measured value of instrument becomes a smoother curve, and meter sensitivity reduces.

Host computer is sampled to the field instrument data of certain hour width, adopts the one-variable linear regression method to carry out analysis and judgement.The slope of one-variable linear regression method institute fitting a straight line has characterized the variation of instrument measurement value trend well, the fine variation that characterizes meter sensitivity of residual sum of squares (RSS).According to expert knowledge library, compare with these two data and historical data, can judge instrument whether meter freezing fault has taken place.

After meter freezing fault occurred, gradual skew took place in the measured value of instrument, can judge the variation of trend and the variable quantity of certain hour width with the slope of one-variable linear regression method institute fitting a straight line.Pairing comparision can be divided into when regulating loop not two kinds of situations when regulating loop:

Instrument is when regulating loop, according to the technological process of this regulating loop, the process measurements of other instrument that the measured value that instrument therewith arranged changes with (or opposite) direction as a reference, when meter freezing fault occurs, situation is just in time opposite, both: the measured value of instrument the variation of (or identical) on the contrary occurred with the variation of (or opposite) direction therewith, surpasses insincere value at certain hour width (data window) variable quantity, and both decidable was a meter freezing fault.

Instrument is not when regulating loop still is with interlocking, according to technological process, the measured value of instrument does not change with the process measurements of other instrument of (or opposite) direction variation therewith, this table surpasses insincere value at certain hour width (data window) variable quantity, and decidable is a meter freezing fault.

After meter freezing fault occurred, gradual skew took place in the measured value of instrument, can judge the variation of trend and the variable quantity of certain hour width with the slope of one-variable linear regression method institute fitting a straight line.

The measured value curve that the back instrument appears in meter freezing fault becomes a smoother curve, and meter sensitivity reduces, and this mainly is that the transmitter bellows freezes institute extremely.At this moment the residual sum of squares (RSS) of the one-variable linear regression method institute fitting a straight line of instrument measurement value sign sensitivity diminishes even levels off to zero, and decidable is a meter freezing fault.

The variation principle of data window: meter freezing fault takes place when hanging down according to outside air temperature, one-variable linear regression method institute fitting a straight line is steeper, otherwise it is then milder, so the variation principle of data window is foundation with the outside air temperature, the data window time can change with outdoor temperature, time weak point when temperature is low, on the contrary then long.

The input that meter freezing fault detects is to drop to＜=be as the criterion-2 ℃ the time with outside air temperature.Here require DCS that one road outside air temperature measured value is arranged

The realization of image data fast: host computer is a slow process by the data acquisition of opc server interface, gathers a secondary data, and judges it is helpless fast for requiring in the fastest 1 minute.In the DCS configuration program, all have a computing module, computing in a second is once.We utilize the characteristic of computing module high-speed sampling computing, carry out pre-service to 1 minute with interior data, and the mean value that calculates 10 6 seconds measured values calls for host computer.By pre-service, realized gathering a secondary data second, meet the needs of high-speed sampling, to satisfy the requirement of judging meter freezing fault fast.

Beneficial effect of the present invention mainly shows: the model that the algorithm that 1, adopts is set up has met better and has frozen the table real process, effective detection alarm after meter freezing fault occurs; 2, rapidity is good; 3, instrument can be reported to the police after the automatic adjustment circuit meter freezing fault occurs in advance, and the operator can fast hired roughneck move operation, and the adjusting process parameter avoids technology big fluctuation to occur; 4, instrument can be reported to the police after the interlock circuit meter freezing fault occurs in advance, and the instrument operator is the bypass interlocking rapidly, takes place to avoid jumping car; 5, host computer monitors constantly whether meter freezing fault takes place, with the particularly tired and carelessness at night of avoiding operating personnel.

Description of drawings

Fig. 1 is the data call FB(flow block).

Embodiment

Below in conjunction with accompanying drawing the present invention is further described.

Embodiment 1

1), the sampling data window time of measurand A is set with reference to Fig. 1, a kind of instrument meter freezing quick fault testing method comprises:, gather n measurand, and each measurand is carried out elimination of rough difference;

2), set up the simple regression linear equation, host computer carries out fitting a straight line to the DCS The data one-variable linear regression method in the sampling data window time, the equation of line of match is (1):

$\hat{A}=\hat{a}+\hat{b}t---\left(1\right)$

Wherein,

$\hat{b}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{t}_i{A}_i-n\stackrel{\‾}{t}\stackrel{\‾}{A}}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{t_i}^2-n{\stackrel{\‾}{t}}^2}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(t_i-\stackrel{\‾}{t})(A_i-\stackrel{\‾}{A})}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(t_i-\stackrel{\‾}{t})}^2}$

$\stackrel{\‾}{t}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{t}_i$ $\stackrel{\‾}{A}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{A}_i$

$\hat{a}=\stackrel{\‾}{A}-\hat{b}\stackrel{\‾}{t}$

Slope

The change direction of expression instrument measurement value trend;

3), adopt pairing comparision to judge instrument fault, comprising:

(3.1), when instrument to be detected at regulating loop, set the measured value of described instrument to be detected and the equidirectional or reverse direction variation of measured value of other instrument, when reverse direction or equidirectional variation appear in the measured value that detects instrument to be detected and other instrument, be judged to be meter freezing fault;

(3.2), when instrument to be detected not at regulating loop, according to the variable quantity in straight-line equation (1) the calculating sampling data window time,, be judged to be meter freezing fault when described variable quantity surpasses upper limit threshold.

In described step 1), the process of each measurand being carried out elimination of rough difference is:

(1.1), to the measurand calculating mean value in the Preset Time section, establish A _iBe i instrument sampled measurement variable, mean value calculation formula (2) is:

$\stackrel{\‾}{A}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{A}_i---\left(2\right)$

Wherein, A: be mean value, n measures total degree in the Preset Time section;

(1.2), represent the error that calculates after the true value, note is made V with arithmetic mean X mean value _i, V _i=A _i-A

Can get the estimated value of standard deviation δ according to the Bezier formula, calculating formula is (3):

$\mathrm{\δ}\≈\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{v}_i^2}{(n-1)}}---\left(3\right)$

(1.3), if certain remainder error | V _i| 3 σ, judge in this measured value and contain gross error, should give rejecting, with A _iAfter the rejecting, after recomputating mean value and standard deviation, the remaining measured value of usefulness judges by above-mentioned steps that still until there not being gross error, elimination of rough difference finishes.

In described step (1), a secondary data was gathered in one second in the every interval of the supercomputing module among the data acquisition system (DAS) DCS, and carried out pre-service to 1 minute with interior data, and the mean value that promptly calculates 10 6 seconds measured values calls for host computer.Described sampling data window time is along with the outdoor temperature rule change: outdoor temperature is low more, and sampling data window time is short more.When outdoor temperature drops to subzero 2 degrees centigrade, drop into described supercomputing module.

The DCS system 1 of present embodiment includes field instrument (comprising transmitter), control station, active station, and DCS system 1 connects DCS computing module 2, and DCS system 1 is connected with opc server 3 with DCS computing module 2 simultaneously, and described opc server 2 is connected with host computer 4.

On-the-spot technology signal is transferred to the DCS control station through field instrument (comprising transmitter), and control station is transferred to DCS active station, opc server by control bus, and opc server is transferred to host computer.Host computer procedure is realized with VB.

The specific implementation algorithm of present embodiment

DCS high-speed sampling algorithm: utilize the characteristic of high-speed sampling, can realize sample of measured value 1 second 1 time, calculated a mean value in per then 6 seconds, 1 minute the conduct of totally 10 numbers upload data, deposit variables A ₁-A ₁₀

Mean value calculation:

If A _iFor i instrument sampled measurement variable then

$\stackrel{\‾}{A}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{A}_i$

A: mean value, n=6

This algorithm software realizes having high real-time at DCS control station CPU.

The host computer algorithm comprises the excluding gross error algorithm, sets up simple regression linear equation and data window set algorithm, specifically has:

One. the excluding gross error algorithm

We utilize 1 minute 10 data (each is 6 seconds average value measured) of computing module high-speed sampling, carry out the excluding gross error pre-service after host computer calls.

A. calculating mean value:

By formula $\stackrel{\‾}{A}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{A}_i$ (n-data number)

B. basis of calculation deviation:

Represent the error that calculates after the true value with arithmetic mean X mean value, be called residual error,

Note is made V _i, V _i=A _i-A

Can try to achieve the estimated value of standard deviation δ according to Bezier (Bessel) formula:

$\mathrm{\δ}\≈\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{v}_i^2}{(n-1)}}---\left(3\right)$

C. judge by 3 σ:

If certain remainder error | V _i| 3 σ, then judge in this measured value and contain gross error, should give rejecting.

D. with A _iReject, judge by above-mentioned steps that still until there not being gross error, elimination of rough difference finishes after the remaining measured value of usefulness recomputates mean value and standard deviation.

Two, simple regression linear equation:

If: A=f (t) A is the instrument measurement value, and it is the function of time t.Some groups of data (t of n point are arranged ₁, A ₁); (t ₂, A ₂) ... (t _n, A _n)

Order $\hat{A}=a+\mathrm{bt}---\left(4\right)$

A is a straight line intercept constant, and b is the straight slope constant.Can be roughly linear relationship.

Determine a with least square method; B.Order

$Q(a,b)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{[A_i-(a+{\mathrm{bt}}_i)]}^2$

Q (a, b) the limit place should satisfy:

$\left\{\begin{array}{c}\frac{\∂Q}{\∂a}=0\\ \frac{\∂Q}{\∂b}=0\end{array}\right.$

Both:

$\left\{\begin{array}{c}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}[A_i-(a+{\mathrm{bt}}_i)]=0\\ \underset{i=1}{\overset{n}{\mathrm{\Σ}}}[A_i-(a+{\mathrm{bt}}_i)]t_i=0\end{array}\right.$

Can get:

$\left\{\begin{array}{c}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{A}_i-\mathrm{na}-b\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t}_i=0\\ \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t}_i{A}_i-a\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t}_i-b\underset{i=1}{\overset{\mathrm{<figure-callout\; id="2"\; label="n\; t\; i"\; filenames="A200810119309E00171.png"\; state="\{\{state\}\}">n</figure-callout>}}{\mathrm{\&Sigma;}}}{t_i}^{}{{}_{}}^2=0\end{array}\right.---\left(5\right)$

Following formula is a normal equations, separates normal equations and gets a, b, is designated as:

(

Be Q (a, minimum point b))

$\hat{b}=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t}_i{A}_i-n\stackrel{\&OverBar;}{t}\stackrel{\&OverBar;}{A}}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t_i}^2-n{\stackrel{\&OverBar;}{t}}^2}=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(t_i-\stackrel{\&OverBar;}{t})(A_i-\stackrel{\&OverBar;}{A})}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(t_i-\stackrel{\&OverBar;}{t})}^2}---\left(6\right)$

Here:

$\stackrel{\&OverBar;}{t}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t}_i$ $\stackrel{\&OverBar;}{A}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{A}_i$

Have:

$\hat{a}=\stackrel{\&OverBar;}{A}-\hat{b}\stackrel{\&OverBar;}{t}---\left(7\right)$

(6) formula is a, the system of linear equations of b, and (a b) has minimal value to Q, so there is one group to separate

Can prove that this group is separated makes Q (a, b) minimalization.

Experimental formula, regression straight line:

With (6), (7) formula is determined

Substitution (1):

$\hat{A}=\hat{a}+\hat{b}t---\left(1\right)$

(1) formula is exactly the experimental formula of t and A correlationship, is called regression beeline equation.

Three, data window set algorithm: T=kw T: data window setting-up time width, k: scale-up factor, w: outdoor temperature value.

Host computer can be realized calling and the data excluding gross error for 1 minute 1 time by the data of opc server interface interchange DCS.To the data in 5 minutes step-lengths, adopt the one-variable linear regression method, return and be straight line, and calculated 5 minutes and 10 minutes increments, according to increment, straight slope etc. and judgement:

1. increasing or decreasing whether.

2. whether automatic adjustment circuit instrument measurement value is with opposite with reference to instrument measurement value increase and decrease.

3. this secondary data is deposited historical data base.

If the data window time is set to 5 minutes.Behind 1 minute data excluding gross error, carry out regressing calculation one time with preceding 4 minutes data, as exist linear dependence then will see the variation of straight slope, according to this as freezing the important evidence that table is judged.Also want judgment data to begin, as freezing the auxiliary foundation that table is judged to the variable quantity at 10 minutes ends.Regressing calculation is to carry out once in per 1 minute, and promptly per minute calculates preceding 5 minutes data call once.

In order to alleviate the DCS computational load, the input of the supercomputing module of DCS is to drop to＜=be as the criterion-2 ℃ the time with outside air temperature.One switch instrument also is set in DCS inside realizes that control artificial input in winter or people are for withdrawing from.

When tested instrument during at automatic adjustment circuit, according to the technological process of this regulating loop, choose 1 to 2 as a comparison with reference to instrument, for example: gasifier nozzle cooling water flow instrument, orifice plate detects, and adopts differential pressure transmitter, generally adds two flow measurement instruments before the burner inlet and after the outlet, two instrument flows equate, differ in error range.Can be used as reference with instrument not at regulating loop, when the regulating loop measuring instrument freezes table, the indication of table can be offset, under the effect of automatic adjustment circuit, variation has taken place in the burner cooling water flow, can make the instrument of regulating loop and another with reference to instrument rightabout skew take place soon like this.It is 1.5% that orifice plate is measured mistake meal, so two epiphase differences can be judged as meter freezing fault greater than 5% of full scale in the certain hour width.

Because during the burner internal leakage, the two is measured phase difference and can increase, but these data can exist and call correction in the historical data base.

When tested instrument not at automatic adjustment circuit, freeze when table, have only this instrument indication itself that skew has taken place and actual process parameter does not change promptly with reference to not variation of instrumented data, deviation can be judged meter freezing fault greater than 5% in the certain hour width.DCS computing module algorithm is identical, and the host computer algorithm does not just have the judgement that reverse direction changes, and is slightly different.

Embodiment 2

With reference to Fig. 1, in the present embodiment, in step 2) in calculate residual sum of squares (RSS):

$Q=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(A_i-{\hat{A}}_i)}^2;$

Described fault detection method also comprises:

4), the historical data the when residual sum of squares (RSS) of institute's fitting a straight line and instrument to be detected are normally moved does the ratio computing, less than threshold value, is judged to be meter freezing fault as ratio.

Host computer can be realized calling and analytical calculation for 1 minute 1 time, to the data in 5 minutes step-lengths by the data of opc server interface interchange DCS, adopt the one-variable linear regression method, it is straight line that data are returned, and calculates 5 minutes and 10 minutes increments, also judges according to increment, straight slope etc.:

4. increasing or decreasing whether.

5. call historical data and check two table inherent errors, residual sum of squares (RSS) just often.

6. calculate two Watch Errors and whether subtract inherent error greater than 5%.

7. with just often residual error duplicate ratio, whether residual sum of squares (RSS) diminishes.

8. this secondary data is deposited historical data base.

Other steps of present embodiment are identical with embodiment 1 with the course of work.

For example: one group of data of a gaging pressure instrument are carried out one-variable linear regression.

Just often: [12.66; 12.33; 12.69; 11.9; 11.8; 12.16; 11.4; 11.3; 12.56; 11.1; 11.3; 11.21; 11.58; 12.09; 12.35; 12.37; 12.53; 12.12; 11.98; 11.65; 11.09; 11.0; 11.26; 11.58; 11.97; 12.08; 12.09; 12.26; 12.42; 12.35; 12.08; 11.9; 12.17; 12.42; 12.15; 11.88; 11.82; 11.87; 11.12; 10.79; 10.74; 10.28; 11.53; 11.81; 10.9; 10.5; 10.3]

Fitting precision: 0.5482, residual sum of squares (RSS) Q:14.7236, variance: 9.3228

When freezing table: [12.78; 12.23; 12.57; 12.84; 12.93; 13.14; 13.03; 12.88; 12.26; 11.83; 11.28; 10.74; 10.33; 9.54; 9.09; 8.76; 8.45; 8.0; 7.75; 7.24; 7.0; 6.79; 6.69; 6.22; 6.35; 5.89; 5.99; 5.57; 5.62; 5.2; 5.02; 4.99; 4.5; 4.27; 4.03; 3.78; 3.73; 3.28; 3.23; 2.77; 2.49; 2.18; 1.98; 1.63; 1.51; 1.0; 0.9; .043; 0.18; 0.03]

Fitting precision: 0.1438, residual sum of squares (RSS) Q:0.868, variance: 6.0397

When as seen freezing the table generation, although bigger skew has taken place in data value, from 12.58 to 0.18, because meter sensitivity reduces, residual sum of squares (RSS) Q value is reduced to 5.9% of normal value, becomes the curve of a smooth decline.Here variance is reduced to 64.8% of normal value.

Claims (6)

1, a kind of instrument meter freezing fault rapid detecting method is characterized in that: described instrument meter freezing fault rapid detecting method comprises:

1), the sampling data window time of measurand A is set, gather n measurand, and each measurand carried out elimination of rough difference;

2), set up the simple regression linear equation, host computer carries out fitting a straight line to the DCS The data one-variable linear regression method in the sampling data window time, the equation of line of match is (1):

$\hat{A}=\hat{a}+\hat{b}t---\left(1\right)$

Wherein,

$\hat{b}=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t}_i{A}_i-n\stackrel{\&OverBar;}{t}\stackrel{\&OverBar;}{A}}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t_i}^2-n{\stackrel{\&OverBar;}{t}}^2}=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(t_i-\stackrel{\&OverBar;}{t})(A_i-\stackrel{\&OverBar;}{A})}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(t_i-\stackrel{\&OverBar;}{t})}^2}$

$\stackrel{\&OverBar;}{t}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{t}_i$ $\stackrel{\&OverBar;}{A}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{A}_i$

$\hat{a}=\stackrel{\&OverBar;}{A}-\hat{b}\stackrel{\&OverBar;}{t}$

Slope

The change direction of expression instrument measurement value trend;

3), adopt pairing comparision to judge instrument fault, comprising:

(3.1), when instrument to be detected at regulating loop, set the measured value of described instrument to be detected and the equidirectional or reverse direction variation of measured value of other instrument, when reverse direction or equidirectional variation appear in the measured value that detects instrument to be detected and other instrument, be judged to be meter freezing fault;

(3.2), when instrument to be detected not at regulating loop, according to the variable quantity in straight-line equation (1) the calculating sampling data window time,, be judged to be meter freezing fault when described variable quantity surpasses upper limit threshold.

2, instrument meter freezing fault rapid detecting method as claimed in claim 1 is characterized in that: in step 2) in calculate residual sum of squares (RSS):

$Q=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(A_i-{\hat{A}}_i)}^2;$

Described fault rapid detecting method also comprises:

4), the historical data the when residual sum of squares (RSS) of institute's fitting a straight line and instrument to be detected are normally moved does the ratio computing, less than threshold value, is judged to be meter freezing fault as ratio.

3, instrument meter freezing fault rapid detecting method as claimed in claim 1 or 2 is characterized in that: in described step 1), the process of each measurand being carried out elimination of rough difference is:

(1.1), to the measurand calculating mean value in the Preset Time section, establish A _iBe i instrument sampled measurement variable, mean value calculation formula (2) is:

$\stackrel{\&OverBar;}{A}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{A}_i---\left(2\right)$

Wherein, A: be mean value, n measures total degree in the Preset Time section;

(1.2), represent the error that calculates after the true value, note is made V with arithmetic mean X mean value _i, V _i=A _i-A

Can get the estimated value of standard deviation δ according to the Bezier formula, calculating formula is (3):

$\mathrm{\&delta;}\&ap;\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{V}_i^2}{(n-1)}}---\left(3\right)$

(1.3), if certain remainder error | V _i| 3 σ, judge in this measured value and contain gross error, should give rejecting, with A _iAfter the rejecting, after recomputating mean value and standard deviation, the remaining measured value of usefulness judges by above-mentioned steps that still until there not being gross error, elimination of rough difference finishes.

4, instrument meter freezing fault rapid detecting method as claimed in claim 3, it is characterized in that: in described step (1), a secondary data was gathered at the every interval of supercomputing module among the data acquisition system (DAS) DCS in one second, and carrying out pre-service with interior data to 1 minute, the mean value that promptly calculates 10 6 seconds measured values calls for host computer.

5, instrument meter freezing fault rapid detecting method as claimed in claim 3 is characterized in that: described sampling data window time is along with the outdoor temperature rule change: outdoor temperature is low more, and sampling data window time is short more.

6, instrument meter freezing fault rapid detecting method as claimed in claim 4 is characterized in that: when outdoor temperature drops to subzero 2 degrees centigrade, drop into described supercomputing module.

Images