CN101858811A - Signal Compensation Method of High Accuracy Pressure Sensor - Google Patents

Abstract

The invention discloses a high-precision pressure sensor signal compensation method. The pressure measurement signal and temperature measurement signal obtained by the pressure sensor are input into a digital signal processor; in the digital signal processor, the original pressure The signal is transformed into: a pressure signal without temperature compensation but eliminating the hysteresis error; through the signal interface processing method, the above pressure signal is temperature corrected to obtain a temperature corrected pressure signal; through the temperature compensation method, the temperature corrected The pressure signal and temperature signal are processed to obtain the pressure signal and temperature signal after temperature compensation and nonlinear error compensation. The invention can compensate hysteresis error, non-linear error and error caused by ambient temperature change of the pressure sensor, and improve the measurement accuracy of the pressure sensor.

Description

Signal Compensation Method of High Accuracy Pressure Sensor

Technical field:

The invention belongs to the field of signal processing, and relates to a sensor signal compensation method, in particular to a compensation method for nonlinear errors, hysteresis errors and errors caused by temperature changes of silicon pressure sensors.

Background technique:

Silicon pressure sensors are the most successful sensor products in micromechanical technology. There are mainly three types: silicon piezoresistive, capacitive and resonant, among which silicon piezoresistive is the most widely used. Silicon piezoresistive pressure sensors are made using the piezoresistive effect of semiconductor material silicon, the principle of Wheatstone bridge, integrated circuit technology and micromachining technology. Silicon piezoresistive pressure sensors have been widely used as miniature vacuum gauges, absolute pressure gauges, flow meters, flow meters, acoustic sensors, pneumatic sensors, etc. Process controllers, etc., are used in cutting-edge technology and industrial fields such as petroleum, chemical industry, biology, medical treatment, aerospace, ocean engineering, and atomic energy.

The static indicators to measure sensor performance mainly include nonlinear error, hysteresis error and repeatability error. In order to improve the measurement accuracy of the sensor, these errors need to be compensated. At present, the compensation method of nonlinear error has been very mature, and the commonly used compensation methods of nonlinear error include look-up table method, curve fitting method and neural network method. Repeatability error is a random error, which needs to be analyzed by statistical methods, and it cannot be compensated at present. Hysteresis is a multi-value corresponding, irregular, and non-smooth special phenomenon, which is caused by energy absorption and transfer delays in the internal components of the sensor. Hysteresis is related to the loading process of the external load to which the sensor is subjected. Because the law of hysteresis error is very complex, there is no report on the application of hysteresis error compensation for silicon pressure sensors. The proportion of hysteresis error to the basic error is usually about 30%, which is an important factor affecting the measurement accuracy of silicon pressure sensors.

The disadvantage of silicon piezoresistive pressure sensors is that they are very sensitive to temperature changes, and their zero point output and sensitivity will change slightly with temperature changes. This phenomenon is called temperature drift. In order to reduce the impact of temperature changes on the measurement accuracy of sensors, it is necessary to compensate the errors caused by temperature changes. At present, the commonly used temperature compensation methods in the industry are: hardware compensation method and software compensation method (computer compensation, microprocessor compensation). Hardware compensation methods include compensation circuit methods such as diodes, triodes, and thermistors. The software compensation method is to use a computer or microprocessor to collect pressure signals and temperature signals, and use digital signal processing technology to compensate the error caused by temperature drift to obtain high-precision pressure signals.

Invention content:

The technical problem to be solved by the present invention is to overcome the deficiency that the hysteresis error cannot be compensated in the measurement of the existing silicon pressure sensor and improve the measurement accuracy of the silicon pressure sensor, to provide a method that can compensate the hysteresis error and at the same time compensate the nonlinear error of the silicon pressure sensor A high-precision signal processing method for errors caused by temperature and temperature changes.

The technical scheme adopted in the present invention is a high-precision pressure sensor signal compensation method, and the method is applied in the intelligent pressure sensor system developed by the author. The system includes a silicon pressure sensor, a signal amplification circuit, and an analog-to-digital conversion circuit (A/D ), DSP data acquisition compensation circuit, interface circuit and industrial control computer; On the described silicon pressure sensor, be respectively connected with signal amplification circuit and analog-to-digital conversion circuit (A/D), signal amplification circuit is connected with analog-to-digital conversion circuit (A/D) simultaneously /D) is connected; the analog-to-digital conversion circuit (A/D) is connected with a DSP data acquisition compensation circuit; the DSP data acquisition compensation circuit is through an interface circuit and an industrial control computer; the interface circuit includes CAN field bus and USB interface; the working process of the system: the temperature signal of the sensor environment, together with the voltage signal through the signal amplification, passes through the analog-to-digital conversion circuit (A/D), converts the analog signal into a digital signal, and then passes through the digital signal processor ( DSP) for digital signal processing to obtain high-precision pressure signals and temperature signals for hysteresis error compensation, temperature compensation and nonlinear error compensation. Finally, the data is transmitted to the industrial control computer through CAN field bus or USB interface.

The signal compensation method of the high-precision pressure sensor follows the steps below:

(1) The silicon pressure sensor measures the pressure measurement signal Vp and the temperature measurement signal Vt; the pressure measurement signal Vp enters the DSP data acquisition compensation circuit after passing through the signal amplification circuit and the A/D conversion circuit; the temperature measurement signal Vt is converted by A/D After the circuit enters the DSP data acquisition compensation circuit;

(2) In the DSP data acquisition compensation circuit, the hysteresis error compensation method is used to convert the pressure measurement signal Vp into the pressure value P' that eliminates the hysteresis error;

(3) In the DSP data acquisition compensation circuit, adopt signal interface processing method to carry out temperature correction to pressure value P ', obtain the pressure signal Vpm after temperature correction;

(4) In the DSP data acquisition compensation circuit, the temperature compensation method is used to obtain the pressure signal P and temperature signal T after temperature compensation and nonlinear error compensation from the temperature corrected pressure signal Vpm and temperature signal Vt.

The hysteresis error compensation method refers to:

或

对压力测量信号Vp的极值序列Vp1、Vp2、…、Vpn进行处理，得到经过迟滞误差补偿的压力信号P’的序列(P′₁，P′₂，...，P′_n)；First, use the extreme value sequence Vp1, Vp2, ..., Vpn of the pressure measurement signal Vp to represent the pressure; secondly, judge whether the pressure is in the loading process (that is, the pressure load increasing process) or the unloading process (that is, the pressure load decreasing process); then use hysteretic inverse model

or

Process the extreme value sequences Vp1, Vp2, ..., Vpn of the pressure measurement signal Vp to obtain a sequence of pressure signals P'(P' ₁ , P' ₂ , ..., P' _n ) after hysteresis error compensation;

Hysteretic inverse model for hysteretic error compensation when pressure is applied during loading for

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 45</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&CenterDot;</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\frac{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 50</span>}}}$

Among them, α _n is the current pressure value P′ _n after hysteresis error compensation, and the current pressure is in the loading process;

Y is the input vector, which is composed of the voltage ΔV _n corresponding to the current extreme pressure increment and the previous extreme pressure P _n-1 , that is, Y=(ΔV _n , P _n-1 ), where ΔV _n =Vp _n -Vp _n-1 ;

Y _i is a support vector, that is, a vector composed of training samples, that is, Y _i = ( _xi (a _i , bi ₎ , _bi ), (i=1, 2, ..., 45); α _i is The weight coefficient (i=1,2,...,45) of the support vector machine obtained through training;

为Hysteretic inverse model for hysteretic error compensation when pressure is unloaded

for

${\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 45</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\frac{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 50</span>}}}$

Among them, b _n is the current pressure value P′ _n after hysteresis error compensation, and the current pressure is in the process of unloading;

Y is the input vector, which is composed of the voltage ΔV _n corresponding to the current extreme pressure increment and the previous extreme pressure P _n-1 , that is, Y=(ΔV _n , P _n-1 ), where ΔV _n =Vp _n -Vp _n-1 ;

Y _i is a support vector, that is, a vector composed of training samples, that is, Y _i = ( _xi (a _i , bi ₎ , a _i ), (i=1, 2, ..., 45);

α _i is the weight coefficient (i=1, 2, . . . , 45) of the support vector machine obtained through training.

The signal processing interface method refers to: using the function model Vpm=f(P', Vt) of the pressure signal Vpm on the pressure P and temperature T, the pressure signal P' and the temperature measurement signal Vt without temperature compensation are processed to obtain the process The temperature-corrected pressure signal Vpm; the function model Vpm=f(P', Vt) is shown in the following formula:

Vpm=-5.4969×10 ^-6 +0.7526×P+0.8192·Vt+4.8869×10 ^-4 ·P ² −0.02361·P·Vt−0.03881·Vt ² .

The temperature compensation method refers to: using the function model P=g(Vpm, Vt) of the pressure P about the pressure signal Vpm-temperature signal Vt and the function model T=q(Vpm, Vt) of the temperature T about the pressure signal Vpm-temperature signal Vt Vt), process the pressure signal Vpm and temperature measurement signal Vt into: high-precision pressure signal P and temperature signal T after temperature compensation and nonlinear error compensation;

The function model P=g(Vpm, Vt) of the pressure P with respect to the pressure signal Vpm-temperature signal Vt:

P＝-117.758+1.335×Vpm+45.134×Vt-0.00129×Vpm ² +0.0477×Vpm·Vt-4.5113×Vt ²

The function model T=q(Vpm, Vt) of temperature T with respect to the pressure signal Vpm-temperature signal Vt:

T=2693.282-1.3888×Vpm-1182.152×Vt+0.00103×Vpm ² +0.2441×Vpm·Vt+130.1434×Vt ² .

The beneficial effects of the present invention are: the hysteresis error of the silicon pressure sensor is effectively compensated, and the non-linear error of the silicon pressure sensor and the error caused by temperature change are compensated at the same time, and the measurement accuracy of the silicon pressure sensor is improved; this is a brand new A digital signal processing method for error compensation of a silicon pressure sensor; the error of a pressure sensor with a total accuracy of 0.2% FS (range) compensated by the method of the invention can be reduced by half.

Description of drawings:

Fig. 1 is a schematic diagram of a Wheatstone bridge of a silicon pressure sensor;

Fig. 2 is the structural diagram of the intelligent pressure sensor system used in the embodiment;

Fig. 3 is a structural diagram of the high-precision pressure sensor signal compensation method of the present invention;

Fig. 4 is the experimental data drawing figure of hysteresis model x (a, b) that the present invention uses;

Fig. 5 is the structure of the support vector machine that the present invention uses;

Fig. 6 is a plotting diagram of experimental data of a silicon pressure sensor about pressure P and temperature T;

Fig. 7 is the input pressure figure when the temperature is 30 ℃ in the hysteresis error compensation experiment of embodiment 1;

Fig. 8 is the experimental result of embodiment 1: the comparison diagram of the error value after hysteresis compensation and uncompensated;

Fig. 9 is the input pressure diagram when the temperature is 65°C in the experiment of hysteresis error compensation and temperature compensation in embodiment 2;

Fig. 10 is the experimental result of embodiment 2: a comparison chart of error values after the compensation of the present invention and only nonlinear error compensation.

Detailed ways:

The present invention is described in further detail below in conjunction with accompanying drawing:

In Figure 1, four force-sensitive resistors of a silicon pressure sensor form a Wheatstone bridge. In order to improve the measurement accuracy of the sensor, the silicon pressure sensor is powered by a constant current source. Since the constant current source is used for power supply, the change of the constant current source voltage at both ends of the bridge A and C reflects the change of the ambient temperature of the sensor, while the output voltage at both ends of the bridge B and D reflects the input pressure. This method uses a pressure sensor A system that can measure pressure and temperature at the same time is usually called a "one bridge, two measurements" system. In this embodiment, the "one bridge and two measurements" system is used, which can reduce the use of temperature sensors, facilitate on-site testing, and save experimental costs. Of course, for the high-precision pressure sensor signal compensation method of the present invention, the "one bridge, two measurements" scheme may not be used, and the temperature analog signal may also be obtained from a temperature sensor installed in the same environment as the silicon pressure sensor.

Fig. 2 is the structure of the intelligent pressure sensor system developed by the author used in the embodiment. The intelligent pressure sensor system is composed of silicon piezoresistive pressure sensor, signal amplification circuit, analog-to-digital conversion circuit, digital acquisition and processing circuit and industrial control computer. The constant current source voltage of the silicon pressure sensor is used as a temperature signal, and passes through the analog-to-digital conversion circuit (A/D) together with the voltage signal amplified by the signal, and the analog signal is converted into a digital signal, and then digitalized by a digital signal processor (DSP). Signal processing to obtain high-precision pressure signals and temperature signals for hysteresis error compensation, temperature compensation and nonlinear error compensation. Finally, the data is transmitted to the industrial control computer through CAN field bus or USB interface. Among them, DSP, as the core of the whole system, is responsible for the functions of each chip operation, data acquisition, digital signal processing and communication.

The working process of the system is: after power on, firstly, the program of the system is initialized; secondly, the DSP query is issued by the industrial computer through the USB or CAN interface acquisition command; if the acquisition command is received, a CPU timer is started, and the timer Collect the pressure signal and temperature signal during the interruption, then perform digital filtering and software compensation; finally, upload the compensated pressure and temperature data to the industrial computer through the USB or CAN interface; after the data is uploaded, the DSP queries the acquisition end command. When the collection end command is reached, the system continues to collect and process signals; if the collection end command is received, the system ends the task.

Fig. 3 is the structure of the signal compensation method of the high-precision pressure sensor, including the hysteresis error compensation method, the signal processing interface method and the temperature compensation method.

或

对压力测量信号Vp的极值序列进行处理，得到经过迟滞误差补偿的压力信号P’。The purpose of the hysteresis error compensation method is to eliminate the hysteresis error generated by the pressure measurement signal Vp during the loading and unloading of the pressure P. The hysteresis error compensation method consists of two parts: the first part is to record the extreme value sequence of the pressure measurement signal Vp during the pressure loading process, because hysteresis is related to the loading process, and the extreme value sequence of the pressure measurement signal Vp records the pressure loading process. The second part is to first judge whether the pressure is in the loading or unloading process, and then use the hysteresis inverse model respectively

or

The extreme value sequence of the pressure measurement signal Vp is processed to obtain the pressure signal P' after hysteresis error compensation.

和

的建模数据。Figure 4 is a plot of experimental data for the hysteresis model x(a,b). Establishing correct and high-precision hysteresis model and inverse model is the key to the hysteresis error compensation program. The hysteresis model of the silicon pressure sensor is based on the calibration experiment data of the pressure P-pressure measurement signal Vp about the hysteresis error. The specific experimental process is as follows: Under the conditions of room temperature 30°C and humidity 56%RH, the experiment is carried out according to the JB/T 10524-2005 machinery industry standard. The experimental instruments mainly include: pressure sensor calibration workbench, constant current source, temperature control box and High precision digital multimeter. The measuring range of the silicon pressure sensor used in the embodiment is 40Mpa, comprehensively consider the influence of the training sample on the fitting accuracy and the complexity of the test experiment, divide 0～40Mpa into 8 equal parts for testing, load from 0Mpa to 5Mpa, write down Output voltage, then reduce the load to 0Mpa, record the output voltage, and calculate x(5,0). Then load to 10Mpa, record the output voltage, reduce the load to 5Mpa, record the output voltage, then reduce the load to 0Mpa, record the output voltage, calculate x(10,5) and x(10,0) respectively, and so on , until 40Mpa, get the output voltage x(a, b) experimental data between extremes. The experimental data of the hysteresis model x(a, b) can be processed to obtain hysteresis inverse models for a and b respectively

and

modeling data.

By performing regression analysis on the experimental data of the hysteresis model x(a, b) and the modeling data of the hysteresis inverse model, the hysteresis model of the silicon pressure sensor and the inverse model for hysteresis error compensation can be obtained. Commonly used regression analysis methods include quadratic surface regression analysis method, neural network and other methods. In order to improve the precision of the model established by the regression analysis and take into account the modeling efficiency, the present invention adopts the method of the support vector machine to perform the regression analysis on the modeling data. Support Vector Machine (SVM for short) is a machine learning algorithm based on statistical learning theory. It is based on the statistical learning theory and the principle of minimum structural risk, and seeks the best compromise between the complexity of the model and the learning ability according to the limited sample information, in order to obtain the best promotion ability of the machine learning algorithm, which can guarantee all The solution obtained is the most global solution. Support vector machines show unique advantages in solving small sample and nonlinear problems.

Fig. 5 is the structure of the support vector machine that the present invention uses, and when the mathematical model of support vector machine carries out data fitting to training sample, express with following formula.

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}$

Among them, Y is the input vector to be tested;

n is the number of support vectors, that is, the number of samples;

Y _i is a support vector, that is, a vector composed of training samples. (i=1,2,...,n);

x(Y) is the output corresponding to Y;

α _i is the Lagrangian multiplier corresponding to the weight coefficient. (i=1,2,...,n);

β is the threshold;

K( _xi , x) is the kernel function of the support vector machine.

Support vector machines have various forms of kernel functions, such as linear kernel functions, polynomial kernel functions, and radial basis kernel functions. The support vector machine regression model of the present invention uses the radial basis kernel function, because the radial basis kernel function produces less deviation. The radial basis kernel function is shown in the following formula

$\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\frac{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="p"\; filenames="HDA0000022442890000022.png"\; state="\{\{state\}\}">p</figure-callout></span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$

Among them, ||YY _i || represents the modulus after taking the difference between the input vector and the support vector;

p is the kernel function parameter, and adjusting p can improve the measurement accuracy of the support vector machine.

Learning parameters: The selection of the kernel function parameter p, the insensitive loss function ε, and the penalty factor C has a great influence on the training efficiency and data fitting accuracy of the support vector machine. In practical application, the parameter determination methods mainly include empirical determination and grid search. The author conducts multiple training and comparisons on the samples, and finally selects the parameters as:

Kernel function parameter p=5;

Insensitive loss function parameter ε=0.0001;

Penalty factor C=1000;

The learning samples composed of modeling data are used as support vectors, and all inputs are made at one time to form a support vector machine, and then each input vector in the training samples is input into the support vector machine in turn for training; based on the training samples and the principle of minimum structural risk, the SVM is solved Structural parameters to minimize the deviation between the output vector and the expected output vector in the training sample, at this point, the training of the support vector machine ends. Finally, the structural parameters of the support vector machine based on the training samples that meet the error requirements are obtained: weight coefficients α ₁ , α ₂ , . . . , α _n and threshold β.

The support vector machine with the above structure and parameters is used to establish a model for hysteresis error compensation as follows:

Hysteretic inverse model for hysteretic error compensation when pressure is applied during loading for

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 45</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\frac{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 50</span>}}}$

Among them, α _n is the current pressure value P′ _n after hysteresis error compensation, and the current pressure is in the loading process;

Y is the input vector, which is composed of the voltage ΔV _n corresponding to the current extreme pressure increment and the previous extreme pressure P _n-1 , that is, Y=(ΔV _n , P _n-1 ), where ΔV _n =Vp _n -Vp _n-1 ;

Y _i is a support vector, that is, a vector composed of training samples, that is, Y _i = ( _xi (a _i , bi ₎ , _bi ), (i=1, 2, ..., 45);

α _i is the weight coefficient (i=1, 2, ..., 45) of the support vector machine obtained through training, and the weight coefficient (α ₁ , α ₂ , ..., α ₄₅ ) =

(-2.6512 -16.9909 -3.8090 62.4542 28.7184 8.5403 -121.7887-70.8702 -26.0235 -12.2707

182.5700 122.7173 64.4205 44.4448 33.8128 -193.8146 -147.9245-90.5564 -68.5294 -60.2854

-27.1669 189.9419 146.2461 112.2237 83.7839 86.4418 64.3222 48.3573-124.0973 -106.7623

-84.5529 -67.1388 -63.9511 -65.9918 -55.2276 -21.8993 72.0294 50.453152.5182 43.0396

40.8106 44.0773 48.0054 27.7502 38.5530)

Threshold β = 0

为Hysteretic inverse model for hysteretic error compensation when pressure is unloaded

for

${\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 45</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\frac{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 50</span>}}}$

Among them, b _n is the current pressure value P′ _n after hysteresis error compensation, and the current pressure is in the process of unloading;

Y is the input vector, which is composed of the voltage ΔV _n corresponding to the current extreme pressure increment and the previous extreme pressure P _n-1 , that is, Y=(ΔV _n , P _n-1 ), where ΔV _n =Vp _n -Vp _n-1 ;

Y _i is a support vector, that is, a vector composed of training samples, that is, Y _i = ( _xi (a _i , bi ₎ , a _i ), (i=1, 2, ..., 45);

α _i is the weight coefficient (i=1, 2, ..., 45) of the support vector machine obtained through training, and the weight coefficient (α ₁ , α ₂ , ..., α ₄₅ ) =

(-2.7459 -11.0461 12.1109 -6.2765 11.6702 -2.7377 -18.6998 35.3271-37.7082 22.8823

-5.8964 15.5362 -11.6501 7.9822 8.0876 -28.7951 59.5846-82.8821 81.8036 -53.0894

21.1136 -30.2184 88.5497 -124.8046 143.7415 -94.2631 36.5565 13.2669-74.1277 216.7344

-415.9337 528.2465 -564.6080 425.8700 -171.0906 43.4731 -18.934981.7430 -173.8384 239.6889

-208.0146 157.9626 -57.5958 8.9726 37.1454)

Threshold β = 0

The purpose of the signal processing interface method is to connect the hysteresis error compensation method and the temperature compensation method, and simultaneously perform temperature correction on the pressure signal that has undergone hysteresis error compensation but has not undergone temperature compensation. The function model Vpm=f(P', Vt) of the pressure signal Vpm on the pressure P and temperature T in the signal processing interface method is based on the pressure measurement signal Vp-temperature signal Vt calibration experiment of the silicon pressure sensor on the pressure P and temperature T Based on the data, obtained by regression analysis method.

FIG. 6 is a plot of experimental data of a silicon pressure sensor with respect to pressure P and temperature T. FIG. The specific process of the experiment is as follows: The experimental instruments mainly include: pressure sensor calibration workbench, constant current source, temperature control box and high-precision digital multimeter, and the experiment is carried out with reference to the JB/T 10524-2005 machinery industry standard. The measuring range of the silicon piezoresistive pressure sensor used in the embodiment is 40Mpa, the silicon piezoresistive pressure sensor is put into the temperature control box, and the temperature is respectively 20 ℃, 30 ℃, 40 ℃, 50 ℃, 60 ℃, 65 Under the condition of ℃, the pressure of the pressure sensor-BD terminal voltage-AC terminal voltage is measured and recorded. The pressure range is 0-40Mpa, and the voltage output value is recorded at nine points of 0Mpa, 5Mpa, 10Mpa, 15Mpa, 20Mpa, 25Mpa, 30Mpa, 35Mpa and 40Mpa. The loading process of the pressure sensor is gradually loaded from 0Mpa to the full scale of 40Mpa, and then gradually decreased from the full scale to 0Mpa. Finally, the experimental data are obtained: pressure P-temperature T-pressure measurement signal Vp-temperature signal Vt. Due to the existence of hysteresis, the pressure measurement signals Vp of the forward and reverse strokes are different at the same temperature and pressure. Therefore, at the same temperature and the same pressure, the pressure signal Vp of the forward and reverse strokes is averaged to obtain the pressure signal Vpm, and the pressure signal Vpm and the pressure P are in a one-to-one mapping relationship.

The modeling data is obtained from the experimental data of the silicon pressure sensor on the pressure P and temperature T: pressure signal Vpm-pressure signal P-temperature signal Vt, by performing regression analysis on these data, the pressure signal Vpm is obtained on the pressure signal P and temperature signal The function model of Vt Vpm=f(P', Vt). In the present invention, use quadratic surface regression analysis to set up the function model Vpm=f(P', Vt) of pressure signal Vpm about pressure signal P and temperature signal Vt, function model is as shown in the following formula:

Vpm＝-5.4969×10 ^-6 +0.7526×P+0.8192 Vt+4.8869×10 ^-4 P ² -0.02361 P Vt−0.03881 Vt ²

Using the function model of the pressure signal Vpm with respect to the pressure signal P and the temperature signal Vt, the temperature-corrected pressure signal Vpm is obtained from the non-temperature-compensated pressure value P' and the temperature signal Vt.

The temperature compensation method is as follows: on the basis of the calibration experimental data of the pressure measurement signal Vp-temperature signal Vt of the pressure P and temperature T of the silicon pressure sensor, the pressure P is established by the quadratic surface regression analysis method about the pressure signal Vpm-temperature signal The function model P=g(Vpm, Vt) of Vt and temperature T are about the function model T=q(Vpm, Vt) of pressure signal Vpm-temperature signal Vt; Utilize pressure P function model P=g(Vpm, Vt) and temperature The T function model T=q(Vpm, Vt), processes the pressure signal Vpm and the temperature signal Vt into: the pressure signal P and the temperature signal T after temperature compensation and nonlinear error compensation.

The temperature compensation method requires the use of a pressure P function model P=g(Vpm, Vt) and a temperature T function model T=q(Vpm, Vt). These function models are all based on the calibration experimental data of the pressure measurement signal Vp-temperature signal Vt of the silicon pressure sensor on the pressure P and temperature T. The pressure measurement signal Vp-temperature signal Vt calibration experiment is exactly the same. From the experimental data: pressure P-temperature T-pressure measurement signal Vp-temperature signal Vt, the modeling data can be obtained respectively: (pressure P-pressure signal Vpm-temperature signal Vt) and (temperature T-pressure measurement signal Vp-temperature signal Vt). Carry out quadratic surface regression analysis method to these experimental data, set up the function model P=g(Vpm, Vt) of pressure P about pressure signal Vpm-temperature signal Vt and temperature T respectively about the function model T of pressure signal Vpm-temperature signal Vt =q(Vpm, Vt), the function model is shown in the following formula.

The function model P=g(Vpm, Vt) of the pressure P with respect to the pressure signal Vpm-temperature signal Vt:

P＝-117.758+1.335×Vpm+45.134×Vt-0.00128×Vpm ² +0.0477×Vpm·Vt-4.5113×Vt ²

The function model T=q(Vpm, Vt) of temperature T with respect to the pressure signal Vpm-temperature signal Vt:

T=2693.282-1.3888×Vpm-1182.152×Vt+0.00103×Vpm ² +0.2441×Vpm·Vt+130.1434×Vt ²

Using the pressure P function model P=g(Vpm, Vt) and the temperature T function model T=q(Vpm, Vt), the pressure signal Vpm and the temperature signal Vt are processed into: the pressure signal P after temperature compensation and nonlinear error compensation and temperature signal T.

The following are two examples for testing the high-precision pressure sensor signal compensation method of the present invention. Embodiment 1: The hysteresis error compensation experiment is an experiment conducted to test the effect of hysteresis error compensation under the condition of constant temperature. Embodiment 2: Hysteresis Error Compensation and Temperature Compensation Experiments are conducted under conditions of temperature changes. The experimental instruments mainly include: piston pressure gauge, constant current power supply, temperature control box and high-precision digital multimeter. The measuring range of the silicon pressure sensor used in the embodiment is 0-40Mpa. Under the conditions of room temperature 26°C and humidity 56%RH, the test is carried out with reference to the JB/T 10524-2005 machinery industry standard.

Example 1: Hysteresis Error Compensation Experiment

In order to test the hysteresis compensation effect of the high-precision pressure sensor signal compensation method of the present invention, the pressure extreme value sequence shown in Figure 7 is used as the input pressure, and the temperature of the temperature control box where the sensor is located is 30°C. The output voltage is used as the input signal to calculate and compare with the signal compensation method of the high-precision pressure sensor of the present invention and the method without hysteresis compensation respectively, and the error comparison is shown in Fig. 8 . Wherein, the solid line is the error of the method of the present invention, and the dotted line is the error of the non-hysteresis compensation method. The error value analysis and comparison are shown in Table 1.

Table 1

Error comparison Error mean root mean square of error After hysteresis error compensation -0.0247 0.0675 Without hysteresis error compensation 0.0505 0.4252

From the comparison of the test results, it can be seen that the error of the pressure value after hysteresis compensation is obviously smaller than the error of the pressure value without hysteresis compensation. Therefore, for the hysteresis error of the silicon pressure sensor, it is effective to use the high-precision pressure sensor signal compensation method of the present invention.

Example 2: Hysteresis Error Compensation and Temperature Compensation Experiments

In order to test the overall compensation effect of the high-precision pressure sensor signal compensation method of the present invention, the pressure shown in Figure 9 is used as the input pressure, and the temperature of the temperature control box where the sensor is located is 65°C. The output voltage is used as the input signal, and the signal compensation method of the high-precision pressure sensor and the nonlinear error compensation method of the present invention are respectively used for calculation and comparison. The error comparison is shown in FIG. 10 . Wherein, the solid line is the error of the method of the present invention, and the dotted line is the error of the nonlinear error compensation method. The error value analysis and comparison are shown in Table 2.

Table 2

Error comparison Error mean root mean square of error After hysteresis, nonlinear error compensation and temperature compensation -0.1006 0.0739 After nonlinear error compensation -0.1396 0.4744

It can be seen from the test results that the error of the pressure value after hysteresis, nonlinear error compensation and temperature compensation is significantly reduced, and the signal compensation method of the high-precision pressure sensor of the present invention is effective.

The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments. It cannot be determined that the specific embodiments of the present invention are limited thereto. Under the circumstances, some simple deduction or replacement can also be made, all of which should be regarded as belonging to the scope of patent protection determined by the submitted claims of the present invention.

Claims (4)

1. a method for compensating signal of high-precision pressure sensor is characterized in that, according to following steps:

(1) silicon pressure sensor measures pressure measurement signal Vp and temperature measurement signal Vt; Pressure measurement signal Vp enters DSP data acquisition compensating circuit successively behind signal amplification circuit and A/D change-over circuit; Temperature measurement signal Vt enters DSP data acquisition compensating circuit behind the A/D change-over circuit;

(2) in DSP data acquisition compensating circuit, adopt the lag error compensation method pressure measurement signal Vp to be converted into the pressure value P of eliminating lag error ';

(3) in DSP data acquisition compensating circuit, adopt the signaling interface disposal route to pressure value P ' carry out temperature correction, obtain through the pressure signal Vpm after the temperature correction;

(4) in DSP data acquisition compensating circuit, adopt temperature compensation, obtain the pressure signal P and the temperature signal T of process temperature compensation and nonlinear error compensation by pressure signal Vpm that passes through temperature correction and temperature signal Vt.

2. method for compensating signal of high-precision pressure sensor according to claim 1 is characterized in that described lag error compensation method is meant:

At first, with extreme value sequence Vp1, the Vp2 of pressure measurement signal Vp ..., Vpn represents pressure; Secondly, judge that pressure is in loading procedure or uninstall process; Utilize sluggish inversion model then respectively

Or

To extreme value sequence Vp1, the Vp2 of pressure measurement signal Vp ..., Vpn handles, and obtains the pressure signal P through the lag error compensation ' sequence P ' ₁, P ' ₂..., P ' _n

When pressure in loading procedure, be used for the sluggish inversion model of lag error compensation For

$a_n=\underset{i=1}{\overset{45}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i\&CenterDot;e^{-\frac{{\left|\right|Y-Y_i\left|\right|}^2}{50}}$

Wherein, a _nFor through the current pressure value P of lag error compensation ' _n, current pressure is in loading procedure;

Y is an input vector, by the current pairing voltage Δ of extreme value pressure increment V _nWith previous extreme value pressure P _N-1Form, i.e. Y=(Δ V _n, P _N-1), Δ V wherein _n=Vp _n-Vp _N-1

Y _iBe support vector, i.e. the vector that constitutes by training sample, i.e. Y _i=(x _i(a _i, b _i), b _i), (i=1,2 ..., 45); α _iFor process is trained the weights coefficient i=1 of the support vector machine that obtains, 2 ..., 45;

The sluggish inversion model that is used for the lag error compensation when pressure in uninstall process is

$b_n=\underset{i=1}{\overset{45}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_i\&CenterDot;e^{-\frac{{\left|\right|Y-Y_i\left|\right|}^2}{50}}$

Wherein, b _nFor through the current pressure value P of lag error compensation ' _n, current pressure is in uninstall process;

Y is an input vector, by the current pairing voltage Δ of extreme value pressure increment V _nWith previous extreme value pressure P _N-1Form, i.e. Y=(Δ V _n, P _N-1), Δ V wherein _n=Vp _n-Vp _N-1

Y _iBe support vector, i.e. the vector that constitutes by training sample, i.e. Y _i=(x _i(a _i, b _i), a _i), i=1,2 ..., 45; α _iFor process is trained the weights coefficient i=1 of the support vector machine that obtains, 2 ..., 45.

3. method for compensating signal of high-precision pressure sensor according to claim 1, it is characterized in that, described signal Processing interface method is meant: utilize pressure signal Vpm about the function model Vpm=f of pressure P and temperature T (P ', Vt), by not temperature compensated pressure signal P ' and temperature measurement signal Vt handle the pressure signal Vpm obtain through temperature correction; Function model Vpm=f (P ', Vt) be shown below:

Vpm＝-5.4969×10 ^-6+0.7526×P+0.8192·Vt+4.8869×10 ^-4·P ²-0.02361·P·Vt-0.03881·Vt ²。

4. method for compensating signal of high-precision pressure sensor according to claim 1, it is characterized in that, described temperature compensation is meant: utilize the function model P=g (Vpm of pressure P about pressure signal Vpm-temperature signal Vt, Vt) and temperature T about the function model T=q (Vpm of pressure signal Vpm-temperature signal Vt, Vt), pressure signal Vpm and temperature measurement signal Vt are treated to: through the high-precision pressure signal P and the temperature signal T of temperature compensation and nonlinear error compensation;

Pressure P about the function model P=g of pressure signal Vpm-temperature signal Vt (Vpm, Vt):

P＝-117.758+1.335×Vpm+45.134×Vt-0.00128×Vpm ²+0.0477×Vpm·Vt-4.5113×Vt ²

Temperature T about the function model T=q of pressure signal Vpm-temperature signal Vt (Vpm, Vt):

T＝2693.282-1.3888×Vpm-1182.152×Vt+0.00103×Vpm ²+0.2441×Vpm·Vt+130.1434×Vt ²。

Images