CN104330137B - Method for detecting quantity of stored grains in granary based on test point pressure values sequence - Google Patents

Abstract

The present invention relates to method for detecting quantity of stored grains in granary based on test point pressure values sequence, arrange a circle pressure transducer on silo bottom surface, each pressure transducer is d with the vertical dimension closest to exterior wall；Detect the output valve of each sensor, according to detection modelCalculate granary storage weight to estimateThe present invention is according to the distribution of silo bottom surface pressure and pressure measurement Variation Features, propose the method for detecting quantity of stored grains in granary of a kind of support vector regression model based on test point pressure values sequence, the present invention is according to corresponding sensor arrangement, and its core technology includes support vector regression quantity of stored grains in granary detection model based on test point pressure values sequence, system calibrating and two parts of modeling method.It is high that proposed method has accuracy of detection, highly versatile, adapts to the feature such as grain storage quantity detection of multiple barn structure type.

Description

Granary grain storage quantity detection method based on detection point pressure value sequence

Technical Field

The invention relates to a granary grain storage quantity detection method adopting a support vector regression model based on a detection point pressure value sequence. Belongs to the technical field of sensor networks.

Background

The grain safety includes quantity safety and quality safety. The online grain quantity detection technology and the system research application are important guarantee technologies for national grain quantity safety, and the development of the research and application on the aspect of national grain safety has important significance and can generate huge social and economic benefits.

Because of the important position of grains in national safety, the on-line detection of the quantity of grain piles is required to be accurate, rapid and reliable. Meanwhile, because the quantity of the grains is huge and the price is low, the cost of the grain pile quantity on-line detection equipment is low, and the detection equipment is simple and convenient. Therefore, the high precision of detection and the low cost of the detection system are key problems which need to be solved in the development of the online detection system for the number of the granaries.

Patents relevant to the present invention include:

(1) the invention discloses a grain warehouse stored grain quantity detection method based on a pressure sensor (patent No. ZL201010240167.7), and the core technology of the invention comprises a calculation model of the grain warehouse stored grain quantity based on the output mean values of pressure sensors on the bottom surface and the side surface of a grain warehouse and a specific system calibration method. The detection system has the remarkable characteristics that the side pressure sensor is used, the number of pressure sensors is large, and the cost of the detection system is high.

(2) The core technology of the invention patent comprises the new methods of compensation of side friction influence based on the square of the output mean value of a bottom surface pressure sensor, a grain pile weight prediction model based on the output mean value of the bottom surface pressure sensor, prediction model modeling based on the grain weight error ratio, rapid system calibration and the like. The method is characterized in that the model is simple, and the average value output by the bottom surface pressure sensor is only used for detecting the grain weight. The model is only suitable for large-scale granaries because the problem of mutual transfer of the side pressure and the bottom pressure is not fully considered.

Disclosure of Invention

The invention aims to provide a granary grain storage quantity detection method based on a support vector regression model of a detection point pressure value sequence, which is a unique detection method.

In order to achieve the above object, the scheme of the invention comprises:

the method for detecting the grain storage quantity of the granary based on the detection point pressure value sequence comprises the following steps:

1) a circle of pressure sensors are arranged on the bottom surface of the granary, and the vertical distance between each pressure sensor and the nearest outer wall is d;

2) detecting the output values of the sensors, assuming s_iPressure sensor measurement of point Q_BL(s_i)，i＝1,...,n_BL，n_BLFor the number of arranged pressure sensors, n is arranged_BLThe pressure sensors on the bottom surface of the granary are numbered, and a sensor sequence Q is formed according to the numbering_BL， $Q_{\mathrm{BL}}=(Q_{\mathrm{BL}}\left(s_1\right),Q_{\mathrm{BL}}\left(s_2\right),...,Q_{\mathrm{BL}}\left(s_{n_{\mathrm{BL}}}\right));$ According to the detection model (9)

$\hat{W}=A_B(\underset{j=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_j\exp (-\mathrm{\γ}{\left|\right|Q_{\mathrm{BL}}-Q_{\mathrm{BL}}^j\left|\right|}^2)+b)---\left(9\right)$

Calculating a detected granary stored grain weight estimateWherein A is_BGamma is a parameter greater than 0, α_jB is a parameter obtained by training the SVM and training samples, α_j≠0；For the corresponding support vector point, j 1.. and l, l are the number of support vectors, and the model parameters are determined by the calibration process.

The calibration method comprises the following steps of pressure sensor calibration: the relationship between the output value of the pressure sensor and the pressure intensity is

Q＝k₀+k₁s(Q) (10)

Wherein Q is the applied pressure; s (Q) is the sensor output value; k is a radical of₀、k₁And the calibration coefficient of the sensor.

Arranging and installing calibrated sensors in more than 6 granaries according to the arrangement mode of the step 1), feeding grains to full granaries and flattening, collecting the output value of the pressure sensor of each granary after the output value of the pressure sensor is stable, and collecting the output value of the pressure sensor according to the calibration coefficient k of each pressure sensor₀、k₁Calculating the pressure value of each sensor, and storing grain quantity W of any given grain bin to be detected according to the serial number of the pressure sensors_mAnd the area of the bottom surface of the detected granaryThe corresponding output value sequence of the bottom surface pressure sensor is expressed as Is the bottom of a granaryNoodle s_iThe pressure value of the point is the corresponding sampleFor various weights of the granary, the sample set is formedWherein M is the number of samples; for a given sample setAnd (3) realizing model modeling shown in the formula (9) by adopting a support vector regression learning algorithm.

And d is 2-4 m.

The invention provides a granary grain storage quantity detection method based on a support vector regression model of a detection point pressure value sequence according to the pressure distribution of the bottom surface of a granary and the change characteristics of pressure measurement values. The method has the characteristics of high detection precision, strong universality, suitability for detecting the quantity of stored grains in various granary structural types and the like.

Drawings

FIG. 1 is a model of a horizontal warehouse floor pressure sensor arrangement;

FIG. 2 is a model of a cartridge floor pressure sensor arrangement;

FIG. 3 is a schematic view of a detection model;

FIG. 4 is a schematic diagram of the specific implementation steps.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings.

The invention discloses a grain quantity detection method, relates to a unique support vector regression model based on a pressure value sequence of detection points, and specifically introduces theoretical premises, corresponding granary sensor arrangement, model derivation, system calibration and modeling and some data obtained by the model in sequence.

1. Theoretical detection model for grain storage quantity of granary

The commonly used grain silos are of the type of horizontal silo, squat silo, silo and the like, after grains are put into the silo, the top of a grain pile is required to be flattened, the shape of the grain pile of the horizontal silo is approximately a cube with different sizes, and the shape of the grain pile of the squat silo and the silo is approximately a cylinder with different sizes. For a given grain bin and grain type, assume that the pressure at point s in the bottom and sides of the grain heap is Q, respectively_B(s)、Q_F(s), the pressure mean values of the bottom surface and the side surface of the grain pile are shown as a formula (1) and a formula (2).

${\stackrel{\‾}{Q}}_B\left(s\right)=\frac{1}{n_B}\underset{i=0}{\overset{n_B}{\mathrm{\Σ}}}{Q}_B\left(s_i\right)---\left(1\right)$

${\stackrel{\‾}{Q}}_F\left(s\right)=\frac{1}{n_F}\underset{j=0}{\overset{n_F}{\mathrm{\Σ}}}{Q}_F\left(s_j\right)---\left(2\right)$

Wherein Q is_B(s_i) Is the bottom surface s of the granary_iA point pressure sensor output value, i ═ 1_B，n_BThe number of the pressure sensors arranged on the bottom surface of the granary; q_F(s_j) Is the side surface s of the granary_jA point pressure sensor output value j 1_F，n_FThe number of the pressure sensors arranged on the side surface of the granary. The stress analysis of the grain stack can be used to obtain that the weight of the grain stack in the granary and the pressure distribution of the granary have the following relationship.

$\hat{W}=A_B({\stackrel{\‾}{Q}}_B\left(s\right)+\frac{C_B}{A_B}{\mathrm{Hf}}_F{\stackrel{\‾}{Q}}_F\left(s\right))---\left(3\right)$

Wherein,for grain bulk weight estimation, A_BIs the area of the bottom of the grain heap C_BThe bottom circumference, H the grain bulk height, f_F(s) is the average coefficient of friction between the side of the grain bulk and the side of the grain bin; Q_B(s)、Q_F(s) is the pressure at point s in the bottom and side of the grain pile, respectively.

For a given grain bin and grain type, the average coefficient of friction f between the sides of the heap and the sides of the bin_FIs constant, the formula (3) can be expressed as

$\hat{W}/A_B=f({\stackrel{\‾}{Q}}_B\left(s\right),{\stackrel{\‾}{Q}}_F\left(s\right),H)---\left(4\right)$

As can be seen from the above formula, the weight of the grain pile and the pressure intensity mean value of the bottom surface of the grain pile are onlyMean lateral pressureAnd the grain bulk height H. Therefore, the core of the granary grain storage quantity detection based on the pressure sensor lies inAnd H, detecting and estimating the three parameters, wherein the grain storage quantity of the granary can be accurately estimated as long as the three parameters are accurately detected and estimated.

According to the average pressure of the side surface of the grain pileAnd the average value of the height H of the grain pile and the pressure intensity of the bottom surface of the grain pileCan be constructed based onIs/are as followsAnd H is estimated to be

${\widehat{\stackrel{\‾}{Q}}}_F\left(s\right)=f_F\left({\stackrel{\‾}{Q}}_B\left(s\right)\right)---\left(5\right)$

$\hat{H}=f_H\left({\stackrel{\‾}{Q}}_B\left(s\right)\right)---\left(6\right)$

Wherein,are respectively asAnd H is estimated.

From the formula (1) to the formula (6), the weight of the grain stored in the granary can be estimatedIs shown as

$\hat{W}/A_B=f(Q_B\left(s_1\right),Q_B\left(s_2\right),...,Q_B\left(s_{n_B}\right))---\left(7\right)$

2. Granary pressure sensor arrangement

According to the distribution characteristics of the pressure of each area of the bottom surface of the granary, the arrangement form of the pressure sensors on the bottom surface of the granary as shown in figures 1 and 2 can be adopted, the number of the pressure sensors is 6-10, the distance between the sensors is not less than 1m, under the condition of ensuring that the grains are convenient to load and unload, the distance d between each pressure sensor and the side wall is as large as possible, and generally about 3 meters can be adopted. In order to ensure the universality of the detection model, the distances d between the pressure sensors of each granary and the side wall are the same, and the number of the pressure sensors is the same.

3. Granary stored grain quantity detection model

According to the relation of the pressure distribution of each region on the bottom surface of the granary, the pressure mean value of the bottom surface of the granary can be approximately estimated by using the pressure mean values of a plurality of detection points on the bottom surface of the granary. Thus, according to the granary floor pressure sensor arrangement shown in fig. 1 and 2, assume s_iPressure sensor measurement of point Q_BL(s_i)，i＝1,...,n_BL，n_BLFor the number of pressure sensors arranged, equation (7) can be approximated as

$\hat{W}/A_B=f(Q_{\mathrm{BL}}\left(s_1\right),Q_{\mathrm{BL}}\left(s_2\right),...,Q_{\mathrm{BL}}\left(s_{n_{\mathrm{BL}}}\right))---\left(8\right)$

N arranged according to a certain order_BLThe pressure sensors on the bottom surface of the granary are numbered, so that a sensor sequence Q can be formed_BL，For any given grain weight W of the grain warehouse to be tested_mAnd the area of the bottom surface of the detected granaryThe corresponding output value of the bottom surface pressure sensor can be expressed as Is the bottom surface s of the granary_iThe pressure value of a point, the corresponding sample can be expressed asMultiple samples may constitute a sample setWhere M is the number of samples.

As the relation model of the grain storage quantity of the granary and the detection value of the pressure sensor shown in the formula (8) has high nonlinearity, and the detection value of the pressure sensor has certain randomness, therefore, the method is based on the sample setModeling may be performed using support vector regression methods.

For a given grain bin, grain type and sample set S, respectivelyValue andis normalized to [ - Δ, Δ ]]Wherein delta is a constant, delta is more than 0 and less than or equal to 2, and a grain storage quantity detection model based on support vector regression can be constructed in the following form by using a common support vector machine model and a training algorithm.

$\hat{W}=A_B(\underset{j=1}{\overset{l}{\mathrm{\Σ}}}{\mathrm{\α}}_j\exp (-\mathrm{\γ}{\left|\right|Q_{\mathrm{BL}}-Q_{\mathrm{BL}}^j\left|\right|}^2)+b)---\left(9\right)$

Wherein A is_BGamma is greater than 0, α_jB is a normal support vector machine model parameter, which can be obtained by using a support vector machine training algorithm, α_j≠0；J is 1, and l is the number of support vectors. FIG. 2 is a schematic diagram of the model shown in formula (9).

And the formula (9) is a support vector regression granary grain storage quantity detection model based on the detection point pressure value sequence. It can be seen that the prediction value of the grain storage quantity of the granary of the detection model depends on As sensor detection values and support vector pointsSupport vector points, distance ofThe detection model is a typical sample point, so the detection model has pattern recognition characteristics based on a template and has good prediction capability.

4. System calibration and modeling method

The system calibration and detection model modeling are carried out according to the following steps:

(1) pressure sensor calibration

In order to ensure the interchangeability of the pressure sensor, the pressure sensor needs to be calibrated for different grain types, so as to obtain the relationship between the output value and the pressure of the pressure sensor as shown in the following formula.

Q＝k₀+k₁s(Q) (10)

Wherein Q is the applied pressure; s (Q) is the sensor output value; k is a radical of₀、k₁And the calibration coefficient of the sensor.

(2) And obtaining system calibration data. By using the sensor arrangement model shown in fig. 1 and 2, calibrated sensors are arranged in more than 6 granaries, the granaries are filled with grains and flattened, after the output value of the pressure sensor is stable, the output value of the pressure sensor of each granary is collected, and the calibration coefficient k of each pressure sensor is used₀、k₁Calculating pressure values of the sensors and forming a sample setWherein, W_mIn order to detect the grain feeding weight of the granary,the area of the bottom surface of the detected granary is M, and the number of samples is M.

(3) And modeling a detection model.

For a given sample setThe model modeling shown in equation (9) can be realized by using a general support vector machine model and a training algorithm. The support vector machine model and the training algorithm belong to conventional technical means, and are not described herein again.

5. Test experiments and results

The length of the horizontal warehouse adopted by the experiment is 9m, the width is 4.2m, and the area is 37.8m². The diameter of the silo is 6m, and the area is 28.26m². According to the pressure sensor arrangement model shown in fig. 1 and 2, for a horizontal warehouse, 15 pressure sensors are arranged along the length direction, and 15 pressure sensors are arranged in a circle in the silo. For each grain tested (wheat, corn and rice), the horizontal warehouse had 6 feeds per test, each about 1 meter and leveled out. The silo feeds grain 8 times in each experiment, and each grain feed is about 1 meter and is flattened.

The method comprises the steps of utilizing 4 times of experimental data of a wheat horizontal warehouse, taking experiments 2, 3 and 4 as modeling samples, utilizing interpolation to form 180 samples, taking experiment 1 data as a test sample, taking a support vector training parameter C to be 3, taking gamma to be 0.02, and obtaining 92 support vector points after training. The results of the calculation of the grain storage weight of each experiment according to the obtained calculation model are shown in tables 1 to 4.

The method comprises the steps of utilizing 4 times of experimental data of a corn silo, taking data of experiments 1, 2 and 3 as modeling samples, utilizing interpolation to form 240 samples, taking data of experiment 4 as a test sample, taking a support vector training parameter C as 3, and taking a support vector training parameter gamma as 0.02, and obtaining 125 support vector points after training. The results of the calculation of the grain storage weight of each experiment according to the obtained calculation model are shown in tables 5 to 8.

The method comprises the steps of utilizing 4 times of experimental data of a corn horizontal warehouse, taking experiments 1, 2 and 3 as modeling samples, utilizing interpolation to form 180 samples, taking experiment 4 data as a test sample, taking a support vector training parameter C to be 3, taking gamma to be 0.02, and obtaining 93 support vector points after training. The results of the calculation of the grain storage weight of each experiment according to the obtained calculation model are shown in tables 9 to 12.

As can be seen from the results of the calculation of the grain storage weight of the granary, the detection results of other detection points are ideal except for the condition of small grain storage weight. Therefore, the grain storage weight detection method is high in measurement accuracy, relatively low in performance requirement on the sensor and suitable for detecting the grain storage quantity of the grain storage bins with various structural types.

One practical detection process is as follows:

the method can be implemented according to the embodiment shown in fig. 4, and the specific steps are implemented as follows:

(1) system configuration

And selecting a specific pressure sensor, and configuring corresponding systems for data acquisition, data transmission and the like.

(2) Bottom surface pressure sensor calibration and installation

Calibrating the pressure sensor according to different grain types to obtain a calibration coefficient k of the pressure sensor₀、k₁. The sensors of the horizontal warehouse are arranged as shown in figure 1, the silo is arranged as shown in figure 2, under the condition of ensuring convenient grain loading and unloading, the distance d between each pressure sensor and the side wall is as large as possible, and the distance can be about 3 meters generally. In order to ensure the universality of the detection model, the distance d between the pressure sensor of each granary and the side wall is the same.

(3) System calibration and modeling

For a given sensor, grain speciesClass and bin type, if the system is not calibrated, arranging calibrated sensors in more than 6 granaries, feeding grains to full granaries and flattening, collecting the output value of the pressure sensor in each granary after the output value of the pressure sensor is stable, and collecting the output value of the pressure sensor according to the calibration coefficient k of each pressure sensor₀、k₁Calculating pressure values of the sensors and forming a sample setWherein, W_mIn order to detect the grain feeding weight of the granary,the area of the bottom surface of the detected granary is M, and the number of samples is M. For a given sample set S, W_mValue andall the values of the components are normalized to [ -delta, delta [ -delta [, ] delta]Wherein delta is a constant, delta is more than 0 and less than or equal to 2, and the model modeling shown in the formula (9) is realized by adopting a common support vector regression learning algorithm.

(4) And (5) detecting the weight of the real bin.

If the system is calibrated, detecting the output of the bottom surface pressure sensor according to the calibration coefficient k of each pressure sensor₀、k₁And calculating the pressure value of each sensor, and detecting the grain storage quantity of the granary by using a model shown in the formula (9).

The specific embodiments are given above, but the present invention is not limited to the described embodiments. The basic idea of the present invention lies in the above basic scheme, and it is obvious to those skilled in the art that no creative effort is needed to design various modified models, formulas and parameters according to the teaching of the present invention. Variations, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the invention, and still fall within the scope of the invention.

Claims (2)

1. The granary grain storage quantity detection method based on the detection point pressure value sequence is characterized by comprising the following steps of:

1) a circle of pressure sensors are arranged on the bottom surface of the granary, and the vertical distance between each pressure sensor and the nearest outer wall is d;

2) detecting the output values of the sensors, assuming s_iPressure sensor measurement of point Q_BL(s_i)，i＝1,...,n_BL，n_BLFor the number of arranged pressure sensors, n is arranged_BLPressure sensor at bottom of granaryThe lines are numbered and a sensor sequence Q is formed according to the numbering_BL，According to the detection model (9)

$\hat{W}=A_B\left(\underset{j=1}{\overset{l}{\Σ}}{\mathrm{\α}}_j\exp \right(-\mathrm{\γ}\left|\right|Q_{BL}-Q_{BL}^j|{|}^2)+b)---\left(9\right)$

Calculating a detected granary stored grain weight estimateWherein A is_BGamma is a parameter greater than 0, α_jB is a parameter obtained by training the SVM and training samples, α_j≠0；To be corresponding toThe support vector points of (1), j, l and l are the number of support vectors, and the model parameters are determined through a calibration process;

the relationship between the output value of the pressure sensor and the pressure intensity is

Q＝k₀+k₁s(Q) (10)

Wherein Q is the applied pressure; s (Q) is the sensor output value; k is a radical of₀、k₁The calibration coefficient of the sensor is obtained;

arranging and installing calibrated sensors in more than 6 granaries according to the arrangement mode of the step 1), feeding grains to full granaries and flattening, collecting the output value of the pressure sensor of each granary after the output value of the pressure sensor is stable, and collecting the output value of the pressure sensor according to the calibration coefficient k of each pressure sensor₀、k₁Calculating the pressure value of each sensor, and storing grain quantity W of any given grain bin to be detected according to the serial number of the pressure sensors_mAnd the area of the bottom surface of the detected granaryThe corresponding output value sequence of the bottom surface pressure sensor is expressed as Q_BL ^m(s_i) Is the bottom surface s of the granary_iThe pressure value of the point is the corresponding sampleFor various weights of the granary, the sample set is formedWherein M is the number of samples; for a given sample setAnd (3) realizing model modeling shown in the formula (9) by adopting a support vector regression learning algorithm.

2. The method for detecting the grain storage quantity of the granary according to claim 1, wherein d is 2-4 meters.