CN102759430B - BP (Back Propagation) neural network based high-precision correction and test method for resonance cylinder pressure sensor - Google Patents

Abstract

The invention belongs to the technical field of test and measurement and particularly relates to a BP (Back Propagation) neural network based high-precision correction and test method for a resonance cylinder pressure sensor, aiming to increase the correction and test precision for the resonance cylinder pressure sensor. The method comprises the steps of: structuring a sensor BP neural network of a dual-implicit-strata network structure, wherein input variables of the network structure are respectively the output period T and the temperature voltage V of the sensor, and output variables of the network structure is the pressure value P; acquiring output periods and temperature voltages of the sensor at different temperature under different pressure input conditions; and collecting output quantities of the sensor at different temperature under different pressure conditions to serve as correction and test as well as inspection sample points. According to the method, with the dual-implicit-strata network structure, the number of network parameters is reduced while the output precision is guaranteed; and the correction and test precision of the resonance cylinder pressure sensor is increased by 25%, and the high-precision correction and test of the sensor is realized.

Description

High-precision Calibration Method of Resonant Cylinder Pressure Sensor Based on BP Neural Network

technical field

The invention belongs to the technical field of testing and measurement, and relates to a high-precision calibration method for a resonant cylinder pressure sensor, in particular to a high-precision calibration method for a resonant cylinder pressure sensor based on a BP neural network.

Background technique

In order to realize the precise control of the missile, the demand for the range and accuracy of the pressure measurement is increasing, so it is necessary to develop a high-precision pressure sensor. For the current resonant cylinder pressure sensor with good stability and high precision, the level of high-precision calibration technology has become an important factor affecting the accuracy of the sensor. In the past, the polynomial fitting method based on the physical model of the sensor is usually used, and the vibration cylinder pressure sensor outputs periodic signals. T and the temperature and voltage signal V, according to the physical model of the sensor, the following multiple nonlinear regression equation can be established between them and the input measurement pressure P:

$P-P_0=\underset{i=0}{\overset{m}{\mathrm{\Σ}}}\underset{j=0}{\overset{\mathrm{no}}{\mathrm{\Σ}}}{K}_{\mathrm{ij}}{(T-T_0)}^i{(V-V_0)}^j$ …………………Formula 1)

In the formula: P——input measurement air pressure;

T——sensor output cycle;

V - pressure sensor temperature voltage;

P ₀ ——reference point air pressure;

T ₀ ——the output period of the reference point sensor;

V ₀ ——reference point temperature voltage;

K _ij ——The characteristic coefficient of the sensor.

There will be a different set of K _ij parameter systems for each vibrating cylinder pressure sensor.

The traditional polynomial fitting technology is based on the formula (1), selects different polynomial orders, and uses the least square method to obtain K _ij coefficients. This method cannot meet the accuracy requirement of the model 0.028%FS in terms of fitting accuracy.

Contents of the invention

The purpose of the invention is to improve the calibration accuracy of the resonant cylinder pressure sensor and provide a high-precision calibration method for the resonant cylinder pressure sensor based on BP neural network.

The technical scheme adopted in the present invention is:

A method for calibrating a pressure sensor of a resonant cylinder based on a BP neural network, including constructing a sensor BP neural network with a double-hidden layer network structure, so that the input variables of the network structure are the output period T and the temperature voltage V of the sensor, and the output variable is the pressure value P; collect the output cycle and temperature voltage of the sensor under different temperature and pressure input conditions; collect the output of the sensor under different temperature and pressure conditions as a calibration test and inspection sample point.

As mentioned above, a resonant cylinder pressure sensor calibration method based on BP neural network, wherein: in the sensor BP neural network, the hyperbolic tangent S-type transfer function is used as the input layer and the first intermediate layer, the first The transfer function between the intermediate layer and the second intermediate layer uses a linear transfer function as the transfer function between the second intermediate layer and the output layer.

A resonant cylinder pressure sensor calibration method based on BP neural network as described above, wherein: when collecting the sample points, multiple temperature points are selected within the range of -45°C to 80°C, and multiple temperature points are selected for each temperature point Sampling point.

A resonant cylinder pressure sensor calibration method based on BP neural network as described above, wherein: in the sensor BP neural network, the network weights of the input layer node and the first intermediate layer node are k coefficients, and the first The network weight of the middle layer node and the second middle layer node is W coefficient, the network weight of the second middle layer node and the output layer node is S coefficient; and the number of nodes in each middle layer is 4; and hyperbolic The tangent sigmoid transfer function is as follows:

${\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 4</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; mi</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

Among them: y _i is the output of the first intermediate layer; x _j is the input variable value; k _ij is the network weight of the input layer and the first intermediate layer; θ _i is the threshold value of each node of the first intermediate layer;

Y _m is the output of the second intermediate layer; W _mi is the network weight of the first intermediate layer and the second intermediate layer; ω _m is the threshold of each node of the second intermediate layer;

The linear transfer function is as follows:

$\mathrm{<span\; class="notranslate">\; o</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 4</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; ln</span>}}{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

Among them: O is the model output; s _ln is the network weight of the second intermediate layer and the output layer; β _l is the threshold of the output layer node.

As mentioned above, a resonant cylinder pressure sensor calibration method based on BP neural network, wherein: the fastest algorithm is used as the training method of BP neural network, and the threshold and weight of each layer are corrected during training, so that the error function along the Descent in the direction of negative gradient.

The beneficial effects of the present invention are:

The proofreading method provided by the present invention uses a double-hidden layer network structure, which reduces the number of network parameters while ensuring the output accuracy; adopts the method of ground training to effectively control the reliability of the training process; and solves the problem by solidifying and initializing the training parameters. The reproducibility of the training process is solved; and the S-type fitting function is constructed, which enhances the approximation ability of the fitting curve and improves the fitting accuracy compared with polynomial fitting.

By constructing a BP neural network model, the present invention increases the calibration accuracy of the pressure sensor of the resonant cylinder by 25%, solves the problem of high-precision calibration of the sensor, and can meet the demand for accurate pressure measurement of missiles.

Description of drawings

Fig. 1 is the sensor BP network model diagram that the present invention adopts;

Figure 2 is a comparison chart of pressure fitting errors.

Detailed ways

A high-precision calibration method for a resonant cylinder pressure sensor based on a BP neural network provided by the present invention is introduced below in conjunction with the accompanying drawings and embodiments:

A method for calibrating a pressure sensor of a resonant cylinder based on a BP neural network, including constructing a sensor BP neural network with a double-hidden layer network structure, so that the input variables of the network structure are the output period T and the temperature voltage V of the sensor, and the output variable is the pressure value P; collect the output cycle and temperature voltage of the sensor under different temperature and pressure input conditions; collect the output of the sensor under different temperature and pressure conditions as a calibration test and inspection sample point.

In the sensor BP neural network, the hyperbolic tangent S-type transfer function is used as the transfer function between the input layer and the first intermediate layer, the first intermediate layer and the second intermediate layer, and the linear transfer function is used as the second intermediate layer. The transfer function between the layer and the output layer.

When collecting sample points, select multiple temperature points within the range of -45°C to 80°C, and select multiple sampling points for each temperature point.

The present invention adopts BP neural network to construct the calibration test model of the resonant cylinder pressure sensor, and the specific contents are as follows:

Firstly, the sample data is collected, and the output cycle and temperature voltage of the pressure sensor are collected under different temperature and pressure input conditions. For example, the test temperature point selects 14 temperature points in the range of -45°C to 80°C. At each temperature point, the pressure measurement points are respectively selected from high to low pressure, a total of 35 pressure points, and the test is recorded after the output signal of the sensor is stable. Data, so that some of them are calibration sample points and some of them are inspection sample points. The number of samples collected should be as large as possible.

In the construction of the sensor BP neural network, the double-hidden layer network structure is selected, so that the input variables of the network structure are the output cycle T and the temperature voltage V of the sensor, and the output variable is the pressure value P; the variables T, V, and P are normalized For the variables X1, X2, O, the BP network model diagram is shown in Figure 1.

Since the determination of the number of layers in the middle layer and the number of nodes in each middle layer has always been a very complicated problem, there is no theoretical guidance, and it often needs to be determined based on experience and multiple experiments. In the present invention, through multiple data simulations, it is determined that the number of intermediate layers is preferably 2 and the number of nodes in each intermediate layer is preferably 4 (for example, 2,4,4,1 structure), the fitting effect is better, and it can also be used other values.

In Figure 1, the network weight of the input layer node and the first intermediate layer node is K coefficient, the network weight of the first intermediate layer node and the second intermediate layer node is W coefficient, and the second intermediate layer node The network weights of the output layer nodes are s coefficients. The network model uses the hyperbolic tangent S-type transfer function as the transfer function between the input layer and the first intermediate layer, and between the first intermediate layer and the second intermediate layer. For the function design, see formula (2) and formula (3):

${\mathrm{the\; y}}_i=\frac{2}{1+e^{-2(\underset{j=1}{\overset{2}{\mathrm{\&Sigma;}}}{k}_{\mathrm{ij}}{x}_j-{\mathrm{\&theta;}}_i)}}-1(i=\mathrm{1,2}...4;j=\mathrm{1,2})$ …… Formula (2)

$Y_m=\frac{2}{1+e^{-2(\underset{i=1}{\overset{4}{\mathrm{\&Sigma;}}}{W}_{\mathrm{mi}}{\mathrm{the\; y}}_i-{\mathrm{\&omega;}}_m)}}-1(m=\mathrm{1,2}...4;i=\mathrm{1,2}...4)$ …… Formula (3)

Among them: y _i is the output of the first intermediate layer; x _j is the input variable value; k _ij is the network weight of the input layer and the first intermediate layer; θ _i is the threshold value of each node of the first intermediate layer. Y _m is the output of the second intermediate layer; W _mi is the network weight of the first intermediate layer and the second intermediate layer; ω _m is the threshold of each node of the second intermediate layer.

The linear transfer function is selected as the transfer function between the second intermediate layer and the output layer, and the function design is shown in formula (4):

$o=\underset{\mathrm{no}=1}{\overset{4}{\mathrm{\&Sigma;}}}{\mathrm{the\; s}}_{\ln }{Y}_{\mathrm{no}}-{\mathrm{\&beta;}}_l(l=1;\mathrm{no}=\mathrm{1,2}...4)$ ………………… Formula (4)

Among them: O is the model output; the output pressure P can be obtained after denormalization; s _ln is the network weight of the second intermediate layer and the output layer; β _l is the threshold of the output layer node.

The sample data is used for model training based on the neural network tool platform. During the training, the threshold and weight of each layer are corrected so that the error function decreases along the negative gradient direction. The values of the network parameters are obtained through sample data training, and one of the training results is as follows:

k _ij =

-3.19283709 -7.16055966

3.07537706 -0.04163307

6.06282238 -9.60802625

-0.99771530 6.42799254

θ _i =

12.46850700

-3.48749525

-2.23594779

-1.30846304

W _mi =

0.06106101 -2.72750488 0.00198993 0.16069464

0.03575073 1.17094160 0.02075128 0.08536263

-0.23359155 0.61221741 1.11657418 0.15811123

0.41401469 0.41214660 1.96806393 0.08391720

ω _m =

2.75440664

1.79626179

-0.94786308

1.49824852

s _ln =

-0.02919945 0.06017406 0.05557906 0.00022849

β _l =

0.88376202

It is preferable to adopt the fastest algorithm (LM algorithm) as the training method of BP neural network, and gradually adjust the parameters such as network learning rate, momentum item, number of iterations and goals to the best state. Using the sample data to carry out model training based on the MATLAB network tool platform, the threshold and weight of each layer are corrected during the training, so that the error function decreases along the negative gradient direction. The value of each network parameter is obtained by training the sample data.

In order to compare the accuracy of the BP neural network fitting model and the polynomial fitting model, calculate the 2 times residual standard deviation 2S _P of the test sample, the polynomial fitting model is 0.250kPa, the BP neural network fitting model is 0.188kPa, and the fitting accuracy is improved by about 25%, as shown in Figure 2. Figure 2 is a comparison effect diagram of pressure measurement errors under different temperature environments. The solid line represents the polynomial fitting pressure measurement error, and the dotted line represents the measurement error of the neural network model.

The invention applies the artificial intelligence neural network to the sensor calibration test, selects the BP neural network to construct the resonant cylinder pressure sensor calibration model, realizes the high-precision pressure output of the sensor, and meets the precise control requirements of the missile.

Claims (2)

1. A kind of resonant tube pressure sensor calibration method based on BP neural network, comprise the sensor BP neural network of construction double hidden layer network structure, make the input variable of network structure be the output cycle T of sensor and temperature voltage V, output variable is Pressure value P; collect the output cycle and temperature voltage of the sensor under different temperature and pressure input conditions; collect the output of the sensor under different temperature and pressure conditions as a calibration test and inspection sample point; In the sensor BP neural network, the hyperbolic tangent S-type transfer function is used as the transfer function between the input layer and the first intermediate layer, the first intermediate layer and the second intermediate layer, and the linear transfer function is used as the second intermediate layer. The transfer function between an intermediate layer and the output layer; When collecting the sample points, multiple temperature points are selected within the range of -45°C to 80°C, and multiple sampling points are selected for each temperature point; In the sensor BP neural network, the network weights of the input layer node and the first intermediate layer node are k coefficients, the network weights of the first intermediate layer node and the second intermediate layer node are W coefficients, and the second The network weights of the intermediate layer nodes and the output layer nodes are s coefficients; and the number of nodes in each intermediate layer is 4; and the hyperbolic tangent S-type transfer function is as follows: ${\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; )</span>$ ${\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 4</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; mi</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$ Among them: y _i is the output of the first intermediate layer; x _j is the input variable value; k _ij is the network weight of the input layer and the first intermediate layer; θ _i is the threshold value of each node of the first intermediate layer; Y _m is the output of the second intermediate layer; W _mi is the network weight of the first intermediate layer and the second intermediate layer; ω _m is the threshold of each node of the second intermediate layer; The linear transfer function is as follows: $\mathrm{<span\; class="notranslate">\; o</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 4</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; ln</span>}}{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$ Among them: O is the model output; s _ln is the network weight of the second intermediate layer and the output layer; β _l is the threshold of the output layer node. 2. a kind of resonant cylinder pressure sensor proofreading method based on BP neural network according to claim 1 is characterized in that: adopt the fastest algorithm as the training method of BP neural network, to the threshold value and the weight of each layer in training Correction is made so that the error function descends in the direction of negative gradient.