CN104197975B - A kind of sensor measuring accuracy raising method based on the constraint of observed value differential - Google Patents

Abstract

The present invention discloses a kind of sensor measuring accuracy raising method based on the constraint of observed value differential, it is determined that iteration times N_int, dynamic conditioning Δt in an iterative process can be realized, adaptively correcting sensor exports, and overcomes the adaptivity that the time delay utilizing arithmetical mean and recurrence average etc. do not possess; The inventive method computing is simple, is applicable to intelligent sensing device self information processing unit and corrects in real time, it is achieved software compensation hardware is not enough, the measurement inaccuracy to a certain degree overcoming sensor self resolving power and causing, it is to increase measuring accuracy; The inventive method, is possible not only to the accuracy improving sensor take off data, also has certain noise removal capability, can be directly used in sensor output data process.

Description

A kind of sensor measuring accuracy raising method based on the constraint of observed value differential

Technical field

The invention belongs to sensor signal processing technology field, more specifically say, relate to a kind of sensor measuring accuracy raising method based on the constraint of observed value differential, in Design of Smart Sensor and signal denoising etc., it is to increase sensor measuring accuracy and removal noise jamming.

Background technology

Along with the arrival of " information age ", as the means of obtaining information, sensor technology obtains significant progress. Sensor application field is more and more extensive, and its requirement is more and more higher, and demand is more and more urgent, and sensor technology has become one of important symbol of a measurement national science state-of-art.

Owing to the signals such as various physical quantity, chemistry amount and biomass can be changed into electrical signal by sensor so that people can utilize computer realize measurement, information processing automatically and automatically control. But the output characteristic of sensor is by the impact of many environmental factorss, such as temperature, power-supply fluctuation, magnetic field etc., sensor exports simultaneously also affects by self operational characteristic, and this causes sensor output often to there is certain error with actual value. In specific application scenario, whole system can be caused bad impact by such error.

For the error that sensor exports with actual value exists, rig-site utilization slip-stick artist also just does some simple filtering and noise reductions for sensor image data (output), this respect also has a lot of ripe algorithm, if kalman filtering, the algorithm such as average that counts are all the algorithms that image data is done post-processed, rarely having for the method that sensor self measuring accuracy improves, it is that the sensor for the particular type of certain applications exists sensor manufacturing process level and explores new measuring method two aspect that major part improves the method for sensor measuring accuracy.

Along with the development of microelectronics and Materials science, sensor is more and more combining in development and apply process with microprocessor, sensor is made not only to have vision, sense of touch, the sense of hearing, the sense of taste, having also had the artificial intelligence such as storage, thinking and Logic judgment ability, the development trend of sensor presents intellectuality. Intelligent sensing device is exactly that a kind of sensor (passing through signal conditioning circuit) having infomation detection and the information processing function concurrently gives intelligent combination with microprocessor. For the design of following intellectualized sensor, utilize the information process unit of sensor self can complete the process of raw measurement data, thus export take off data more accurately.

The using method that can be applied in raw measurement data process at present has the methods such as arithmetic filtering, recurrence average filtering, weighting filtering, and these methods all exist some drawbacks. Although arithmetic filtering can remove random disturbance, but cause take off data time delay big along with sampling point increase. Recursive Filtering and weighting Recursive Filtering, although each calculates moment can return measurement data, but filtered version is single, parameter adjustment blindly, does not have stronger adaptivity, therefore, wish a kind of so simple and dynamic inflation method, the process of some simple sensor raw measurement datas can be completed on this limited calculating unit of sensor self information treating part, namely utilize the method for software compensation hard ware measure defect, thus improve the measuring accuracy of sensor self.

Summary of the invention

It is an object of the invention to for the problem that sensor self resolving power causes sensor measuring accuracy not high, a kind of sensor measuring accuracy raising method based on the constraint of observed value differential is provided, to a certain degree compensating hardware deficiency, according to observed value information, revise observed value, approach ideal value, to improve measuring accuracy.

For realizing above object, the sensor measuring accuracy raising method that the present invention retrains based on observed value differential, it is characterised in that, comprise the following steps:

(1), determine to calculate iteration times N_int;

(2), sensor raw measurement data is obtained, when sampling instant is N_intIndividual sampling instant, starts trimming process, now, and current time i=N_int;

(3) and make j=i-1, the initial generalized velocities of calculating sensor current time i:

$V_{ij}^R=\frac{y_i-y_j}{(i-j)T_s}$

Wherein, T_sFor the sampling time of sensor, y_i,y_jFor the observed value of sampling instant i, j;

(4), j=i-2 is made;

(5), calculating sensor observed value generalized velocities:

$V_{ij}=\frac{y_i-y_j}{(i-j)T_s};$

(6) judge:

If V_ij*V_i(j-1)< 0, the generalized velocities of current time iFor:Otherwise:

If a is V_ij>=0, then:

$V_{ij}^R=\left\{\begin{array}{cc}{V}_{ij};& V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)<V_{ij}<V_{i(j-1)}^R+S_{Verr}(T_s*(i-j\left)\right)\\ V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s);& V_{ij}\&le;V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)\\ V_{i(j-1)}^R+S_{Verr}\left(\right(i-j)*T_s);& V_{ij}\&GreaterEqual;V_{i(j-1)}^R+S_{Verr}\left(\right(i-j)*T_s)\end{array}\right.;$

If $V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)<V_{ij}<V_{i(j-1)}^R+S_{Verr}(T_s*(i-j\left)\right),$ Then make j=j-1, and return step (5), otherwise, enter step (7);

B), V is worked as_ijDuring < 0

$V_{ij}^R=\left\{\begin{array}{cc}{V}_{ij};& V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)<V_{ij}<V_{i(j-1)}^R+S_{Verr}(T_s*(i-j\left)\right)\\ V_{i(j-1)}^R+S_{Verr}\left(\right(i-j)*T_s);& V_{ij}\&le;V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)\\ V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s);& V_{ij}\left(i\right)\&GreaterEqual;V_{i(j-1)}^R+S_{Verr}\left(\right(i-j)*T_s)\end{array}\right.$

If $V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)<V_{ij}<V_{i(j-1)}^R+S_{Verr}(T_s*(i-j\left)\right),$ Then make j=j-1, and return step (5), otherwise, enter step (7);

Wherein: $S_{Verr}\left(\right(i-j)*T_s)=\frac{M}{(i-j)*T_s};$

S_Verr((i-j)*T_s) it is generalized velocities error, got two sampling point timed interval �� t=(i-j) * T_sFunction, M is the resolving power of sensor;

(7), generalized velocities is calculatedAs the generalized velocities after i-th sampling instant place sensor calibration, and according to formula:

$y_{i+1\_correct}=y_i+V_{ij}^R*T_s$

The correction obtaining subsequent time i+1 exports y_{i+1_correct}As the output value of sensor;

(8), current time i add 1, return step (3), ask for N like this_intEach real-time correction value output of sampling instant sensor after individual sampling instant.

The object of the present invention is achieved like this.

This scheme proposes a kind of sensor measuring accuracy raising method based on the constraint of observed value differential, it is determined that iteration times N_int, dynamic conditioning �� t in an iterative process can be realized, adaptively correcting sensor exports, and overcomes the adaptivity that the time delay utilizing arithmetical mean and recurrence average etc. do not possess; The inventive method computing is simple, is applicable to intelligent sensing device self information processing unit and corrects in real time, it is achieved software compensation hardware is not enough, is to a certain degree overcoming what sensor self resolving power caused measurement inaccuracy, it is to increase measuring accuracy; The inventive method, is possible not only to the accuracy improving sensor take off data, also has certain noise removal capability, can be directly used in sensor output data process.

Accompanying drawing explanation

Fig. 1 is the comparison diagram of linear sensing device idealized characteristic and actual characteristic curve;

Fig. 2 is the actual output of stochastic inputs lower sensor and the desirable graphic representation exported;

Fig. 3 is that Fig. 2 middle ideal exports and actual output graph of errors;

Fig. 4 is superimposed with output value and graph of errors thereof after the sinusoidal input lower sensor idea output of white Gaussian noise, real output value and correction;

Fig. 5 is a kind of embodiment schema of sensor measuring accuracy raising method that the present invention retrains based on observed value differential.

Embodiment

Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described, so that the technician of this area understands the present invention better. Requiring particular attention is that, in the following description, when perhaps the detailed description of known function and design can desalinate the main contents of the present invention, these descriptions will be ignored here.

First with the linear sensing device input-output characteristic curve of Fig. 1, wherein M represents sensor intrinsic resolution, and Xres represents sensor input internal discrete sampling interval, and K is linear sensing device ratio system.

In Fig. 1, strokes and dots straight line is ideal output characteristic, and due to so input and output characteristic accurately unavailable in the characteristic reality by sensor process and digitizing, dashed curve is discrete sampling input and output characteristic, is sensor actual measured property. Therebetween difference shows desirable true to certainly exist error between value and observed value, and based on this, under given list entries, ideal value and observed value and error thereof are as shown in Figure 2.

In fig. 2, linear Proportional coefficient K=1, the sensor of intrinsic resolution M=0.1, under given stochastic inputs sequence, obtaining desirable output and measure exporting and the two deviation curve, wherein dotted line is desirable output, and solid line is that actual output is according to observed value information. As shown in Figure 3, observe both deviations of discovery and it is less than intrinsic resolution M. Utilize the thought of dynamic compensation like this, obtain revising observed value, approach ideal value, be specially:

1, the measurement value sensor generalized velocities in sensor discrete sampling situation is defined, that is:

$\frac{\mathrm{\&Delta;}y}{\mathrm{\&Delta;}t}=\frac{y_i-y_j}{(i-j)T_s}---\left(1\right)$

Wherein y_i,y_jFor the observed value of sampling instant i, j, i �� j.

On this basis, defining measurement value sensor generalized velocities error is

Wherein M is sensor intrinsic resolution, and �� t is the timed interval of twice measurement, and proves: in the moment i instantaneous generalized velocities of sensor and calculating generalized velocities, meet: V_R(i)��V_C(i)��S_Verr(�� t), wherein V_R(i) and V_CI () is respectively moment i instantaneous generalized velocities and calculates generalized velocities, V_CI () calculates by (1) formula, V_R(i) for exist in theory but the actual instantaneous generalized velocities that can not ask.

2, propose on self-defined generalized velocities basis and pass through iteration, constantly increase �� t, realize dynamically reducing generalized velocities error, make to calculate generalized velocities and approach instantaneous generalized velocities gradually, until iteration terminates, obtain correction generalized velocities, according to the correction anti-correction method separating observed value of generalized velocities, realize the measurement update to sensor, thus improve measuring accuracy.

The lower calibration result of sinusoidal input is as shown in Figure 4. Wherein dotted line is the desirable output being superimposed with white Gaussian noise, and dotted line is actual output curve, and solid line is the curve after correction. It is 0.01 white Gaussian noise that the sinusoidal Signal averaging being wherein input as 5+sin (t) has average to be 0 variance. The resolving power M of sensor is 0.1. Exporting in application condition, graph of errors when solid line is do not correct, dotted line exports graph of errors for correcting to export with desirable, relatively find, exporting error after correction and be significantly lower than non-school timing error, repeatedly test simultaneously, under ramp input, error average and variance are as shown in table 1 below.

Table 1

(2) on linear sensor input and output characteristic basis, complete the theoretical proof of this correction method validity, and give after the method applies, output value after correction and between idea output statistical error average beWherein N_intFor setting iteration number of times, actual iteration number of times when i-j represents that iteration terminates.

As shown in Figure 5, it is consistent with summary of the invention for the schema of the present invention, comprises and arranges iteration times N_int, obtain sensor raw measurement data, get N_intIndividual raw measurement data, just starts correction. Iteration times N is set_intDetermine according to particular case, as long as can ensure that iteration can terminate. This two step this corresponding to step (1), (2). It should be noted that, sensor raw measurement data constantly carries out, and correction also constantly carries out, and in step (8), current time just can be updated to subsequent time like this.

After starting correction, adopt the generalized velocities after step (3)��(6) iterative computation sensor calibration, then obtain correction with the generalized velocities integration after this sensor calibration and export, this corresponding step (7), final updating sampling instant, and return and calculate N_intEach real-time correction value output of sampling instant sensor after individual sampling instant, thus realize correction and export, until sensor terminates correction.

Although above the embodiment of the present invention's explanation property being described; so that those skilled in the art understand the present invention; but should be clear; the invention is not restricted to the scope of embodiment; to those skilled in the art; as long as various change is in appended scope and the spirit and scope of the present invention determined, these changes are apparent, and all utilize the innovation and creation of present inventive concept all at the row of protection.

Claims (1)

1. the sensor measuring accuracy raising method based on the constraint of observed value differential, it is characterised in that, comprise the following steps:

(1), determine to calculate iteration times N_int;

(2), sensor raw measurement data is obtained, when sampling instant is N_intIndividual sampling instant, starts trimming process, now, and current time i=N_int;

(3) and make j=i-1, the initial generalized velocities of calculating sensor current time i:

$V_{ij}^R=\frac{y_i-y_j}{(i-j)T_s}$

Wherein, T_sFor the sampling time of sensor, y_i,y_jFor the observed value of sampling instant i, j;

(4), j=i-2 is made;

(5), calculating sensor observed value generalized velocities:

$V_{ij}=\frac{y_i-y_j}{(i-j)T_s};$

(6) judge:

If V_ij*V_i(j-1)< 0, the generalized velocities of current time iFor:Otherwise:

If a is V_ij>=0, then:

$V_{ij}^R=\left\{\begin{array}{cc}{V}_{ij};& V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)<V_{ij}<V_{i(j-1)}^R+S_{Verr}(T_s*(i-j\left)\right)\\ V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s);& V_{ij}\&le;V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)\\ V_{i(j-1)}^R+S_{Verr}\left(\right(i-j)*T_s);& V_{ij}\left(i\right)\&le;V_{i(j-1)}^R+S_{Verr}\left(\right(i-j)*T_s)\end{array}\right.;$

If $V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)<V_{ij}<V_{i(j-1)}^R+S_{Verr}(T_s*(i-j\left)\right),$ Then make j=j-1, and return step (5), otherwise, enter step (7);

B), V is worked as_ijDuring < 0

$V_{ij}^R=\left\{\begin{array}{cc}{V}_{ij};& V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)<V_{ij}<V_{i(j-1)}^R+S_{Verr}(T_s*(i-j\left)\right)\\ V_{i(j-1)}^R+S_{Verr}\left(\right(i-j)*T_s);& V_{ij}\&le;V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)\\ V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s);& V_{ij}\left(i\right)\&le;V_{i(j-1)}^R+S_{Verr}\left(\right(i-j)*T_s)\end{array}\right.$

If $V_{i(j-1)}^R-S_{Verr}\left(\right(i-j)*T_s)<V_{ij}<V_{i(j-1)}^R+S_{Verr}(T_s*(i-j\left)\right),$ Then make j=j-1, and return rapid (5), otherwise, enter step (7);

Wherein: $S_{Verr}\left(\right(i-j)*T_s)=\frac{M}{(i-j)*T_s};$

S_Verr((i-j)*T_s) it is generalized velocities error, got two sampling point timed interval �� t=(i-j) * T_sFunction, M is the resolving power of sensor;

(7), generalized velocities is calculatedAs the generalized velocities after i-th sampling instant place sensor calibration, and according to formula:

$y_{i+1\_correct}=y_i+V_{ij}^R*T_s$

The correction obtaining subsequent time i+1 exports y_{i+1_correct}As the output value of sensor;

(8), current time i add 1, return step (3), ask for N like this_intEach real-time correction value output of sampling instant sensor after individual sampling instant.