CN102795627B - Multi-parameter online monitoring and optimizing control device and method of polycrystalline silicon reduction furnace - Google Patents

Abstract

The invention discloses a multi-parameter online monitoring and optimizing control device and method of a polycrystalline silicon reduction furnace. The device comprises a multi-parameter infrared monitoring probe, an infrared image processing and vision measuring module and a process optimizing control module, wherein the multi-parameter infrared monitoring probe obtains an infrared image of a silicon rod in the furnace; the infrared image is collected by a data collecting module and is converted into a digital image; the digital image is analyzed and processed by an image processing module, and the diameter and the growing rate of the silicon rod are obtained through a vision measuring technology; a colorimetric temperature measurement method is used for measuring temperature distribution on the surface of the silicon rod and obtained data is accessed to a display through a user interface; and an optimizing control module is established through measured data obtained by analyzing, and different polycrystalline silicon reduction furnace types are combined to carry out closed ring optimizing control. According to the multi-parameter online monitoring and optimizing control device and method, a silicon rod growing process can be optimally controlled; and the device and the method are of great importance of saving energy and reducing consumption, improving the production efficiency, guaranteeing the production safety and reducing the labor intensity.

Description

Polycrystalline silicon reducing furnace many reference amounts on-line monitoring and optimized control device and method

Technical field

The present invention relates to Si reduction stove production field, be specifically related to a kind of polycrystalline silicon reducing furnace many reference amounts on-line monitoring and optimized control device and method.

Technical background

Polysilicon is the basic material of electronic industry and solar energy industry, is widely used in semi-conductor chip, high-performance sensors, optical fiber, solar panel etc.To 2010, global polysilicon output reached 120,000 tons, and China has accounted for 50% of ultimate production, nearly 40,000,000,000 Renminbi of the output value.Estimate 2011, global polysilicon output will reach 160,000 tons, and Chinese proportion will be brought up to more than 60% of ultimate production.

In reduction furnace, the parameters such as the temperature distribution on silicon rod surface, excellent footpath, growth rate are the key links that improves the silicon rod growth quality.In the polycrystalline silicon rod process of growth, increase in temperature, the sedimentation rate of silicon on the silicon core increases, and the silicon rod growth is accelerated, but power consumption is too much, and temperature is higher, and gas phase condition is harsher, and the homogeneity of silicon rod is poorer; Temperature reduces, and the sedimentation rate of silicon is slack-off, and when temperature, during lower than certain value, silicon rod easily ruptures, the proceeding of impact growth.In addition, the diameter of silicon rod is the important evidence of temperature regulation and feed gas component proportion over time, and irrational proportioning can make the silicon rod growth velocity descend, and consumes in a large number the energy.Therefore, just in the urgent need to a kind of can be in reduction furnace these parameters of on-line monitoring carry out the equipment of closed optimized control.

But the on-line measurement of silicon rod many reference amounts and theoretical investigation that optimal control is all carried out and exploratory test, do not have ripe plant and instrument and enter production application at present both at home and abroad to reduction furnace.In the world, even all there is no the equipment of ripe on-line monitoring and control in the Siemens in the technological precedence status, Mitsubishi Electric Co. etc. aspect the production of polysilicon technology controlling and process yet; Blank especially at home, at present, on-line monitoring and the optimal control of the many reference amounts (as silicon rod diameter, temperature etc.) in reduction furnace rest on theoretical research stage basically.

So far, in the actual production of the whole nation over thousands of seat polycrystalline silicon reducing furnace, what adopt is all that to preset empirical curve be basic open loop type master mode, temperature is by passing through the some thermometric of observation window, fan out from point to area, and the measurement of diameter is estimated roughly by the experience range estimation of workman's work fully, obvious this method has certain subjectivity and randomness.

Given this, developed a kind of equipment that can in reduction furnace, to parameters such as the temperature distribution on silicon rod surface, excellent footpath, growth rates, carry out on-line measurement and carry out closed optimized control.

Summary of the invention

The invention solves the many reference amounts Real-Time Monitoring of polycrystalline silicon growth process and the technical barrier of optimal control, developed the online Measurement accuracy of the parameters such as a kind of silicon rod surface temperature distribution, silicon rod diameter, growth rate and carried out the equipment of closed optimized control.To saving energy and reduce the cost, enhance productivity, ensure production safety, reducing labor intensity all has very important meaning.

The technical solution used in the present invention is:

Polycrystalline silicon reducing furnace many reference amounts on-line monitoring and optimized control device, it is characterized in that, include many reference amounts infrared monitoring probe, infrared image processing and vision measurement module, process optimization control module, described many reference amounts infrared monitoring probe includes near-infrared optical system, image capture module, the water-cooled shield cap is equipped with in many reference amounts infrared monitoring probe outside, the front end of water-cooled shield cap is provided with water-cooling type and seals withstand voltage observation window, and many reference amounts infrared monitoring probe is deep in polycrystalline silicon reducing furnace, described infrared image processing and vision measurement module include infrared image processing module and vision measurement module, the infrared image processing module is connected data acquisition module successively, image processing module, the vision measurement module includes the color comparison temperature measurement module, the diameter measurement module, the output terminal of image processing module respectively with the color comparison temperature measurement module, the diameter measurement module connects, the color comparison temperature measurement module, the equal access data output interface of the output terminal of diameter measurement module, user interface, data output interface is circumscribed with the process optimization control module, the user interface external-connection displayer, the process optimization control module is connected with former Controlling System by a data interface.

Described process optimization control module is the closed optimized control model.

Described image capture module is the infrared CCD camera.

The method of polycrystalline silicon reducing furnace many reference amounts on-line monitoring and optimal control, is characterized in that, comprises following concrete steps:

1) many reference amounts infrared monitoring probe is by choosing optimum infrared band and highly sensitive infrared CCD camera, obtain the picture signal of silicon rod growth in stove by near-infrared optical system, infrared CCD camera, and carry out the A/D conversion through data acquisition module, obtain the infrared digital image of polysilicon rod;

2) infrared digital image of above-mentioned acquisition is sent into to image processing module, and carry out the image pre-treatment and cut apart through image processing module, in conjunction with concrete image, by Image denoising algorithm and level and smooth filtering algorithm, protected to greatest extent the imaging surface Characteristics of The Distribution of Temperature; Find out the edge of sealing by carrying out the profile tracking, and carry out sub-pixel rim detection and fitting of a straight line, obtain the image border of degree of precision;

3) image of above-mentioned processing is transported in the diameter measurement module of vision measurement module, at first the demarcation of carrying out the infrared CCD camera obtains the inside and outside parameter of camera, carry out choosing of primary objective image border according to the Hough Algorithm of fitting a straight line, then image is carried out the tracking at edge, zonule; Under instructing, the fundamental matrix of Robust estimation carries out corresponding point matching; Finally, according to above algorithm, the pixel value of measurement and effective unit yardstick processed are converted, obtained the real-time diameter data of silicon rod;

4) image of processing in step 2 is transported in the color comparison temperature measurement module, utilize the image processing system in the color comparison temperature measurement module image to be carried out to the correction of bad point, to Image Segmentation Using, level and smooth, layering, gradation conversion, then calling data storehouse, grey scale curve is carried out to matching, temperature is demarcated, completed thermometric;

5) by the silicon rod diameter and the surface temperature data that obtain in step 3,4, by data output interface, be transported in the process optimization control module, in conjunction with work information, set up optimizing control models, with feedback, instruct silicon rod to be grown under the most rational processing condition;

6) the silicon rod diameter of above-mentioned acquisition and surface temperature data are outputed to terminal, obtain real-time diameter and the growth rate of silicon rod, and the real time temperature of silicon rod surface any point, and show in indicating meter.

Principle of work of the present invention is:

The present invention obtains silicon rod infrared image in stove by many reference amounts infrared monitoring probe (NICCD), analyzed, processed by image processing module again through data collecting module collected, after being converted to digital picture, obtain silicon rod diameter and growth rate by the vision measurement technology, adopt color comparison temperature measurement method to measure the temperature distribution on silicon rod surface, the take off data that the data obtained obtains by analysis through user interface access indicating meter, set up optimizing control models, in conjunction with different polycrystalline silicon reducing furnace types, carry out closed optimized control.

Beneficial effect of the present invention is:

1) the present invention compares with existing temperature measuring equipment and has that temperature-measuring range is large, precision is high, the advantages such as long-term continuously measured;

2) the present invention compares with current artificial experience estimation silicon rod diameter and growth rate, and having advantages of can be continuously, accurately measurement;

3) the present invention compares with existing fixedly master mode, has and can, for the different type of furnaces and operating mode, provide the advantage of corresponding closed optimized control model.

The accompanying drawing explanation

Fig. 1 is system architecture functional block diagram of the present invention.

Fig. 2 optimizing control models schematic diagram of the present invention,

Fig. 3 is surface temperature monitoring result schematic diagram in the silicon rod process of growth.

Fig. 4 is silicon rod diameter monitoring result schematic diagram in the silicon rod process of growth.

Embodiment

While doing furtherly below in conjunction with example and accompanying drawing to the present invention, but should not limit the scope of the invention with this.

Structure of the present invention forms as shown in Figure 1.

Polycrystalline silicon reducing furnace many reference amounts on-line monitoring and optimized control device, it is characterized in that, include many reference amounts infrared monitoring probe 1, infrared image processing and vision measurement module 2, process optimization control module 3, many reference amounts infrared monitoring probe 1 includes near-infrared optical system 4, image capture module 5, the water-cooled shield cap is equipped with in many reference amounts infrared monitoring probe outside, the front end of water-cooled shield cap is provided with water-cooling type and seals withstand voltage observation window, and many reference amounts infrared monitoring probe is deep in polycrystalline silicon reducing furnace, infrared image processing and vision measurement module 2 include infrared image processing module and vision measurement module, the infrared image processing module is connected data acquisition module 6 successively, image processing module 7, the vision measurement module includes color comparison temperature measurement module 8, diameter measurement module 9, the output terminal of image processing module 7 respectively with color comparison temperature measurement module 8, diameter measurement module 9 connects, color comparison temperature measurement module 8, the equal access data output interface 12 of the output terminal of diameter measurement module 9, user interface 10, data output interface is circumscribed with process optimization control module 3, user interface external-connection displayer 11, process optimization control module 3 is connected with former Controlling System by a data interface.

Process optimization control module 3 is the closed optimized control model.

Image capture module 5 is the infrared CCD camera.

The method of polycrystalline silicon reducing furnace many reference amounts on-line monitoring and optimal control comprises following concrete steps:

1) many reference amounts infrared monitoring probe 1 is by choosing optimum infrared band and highly sensitive infrared CCD camera, obtain the picture signal of silicon rod growth in stove by near-infrared optical system 4, infrared CCD camera, and carry out the A/D conversion through data acquisition module 6, obtain the infrared digital image of polysilicon rod;

2) infrared digital image of above-mentioned acquisition is sent into to image processing module 7, and carry out the image pre-treatment and cut apart through image processing module 7, in conjunction with concrete image, by Image denoising algorithm and level and smooth filtering algorithm, protected to greatest extent the imaging surface Characteristics of The Distribution of Temperature; Find out the edge of sealing by carrying out the profile tracking, and carry out sub-pixel rim detection and fitting of a straight line, obtain the image border of degree of precision;

3) image of above-mentioned processing is transported in the diameter measurement module 9 of vision measurement module, at first the demarcation of carrying out the infrared CCD camera obtains the inside and outside parameter of camera, carry out choosing of primary objective image border according to the Hough Algorithm of fitting a straight line, then image is carried out the tracking at edge, zonule; Under instructing, the fundamental matrix of Robust estimation carries out corresponding point matching; Finally, according to above algorithm, the pixel value of measurement and effective unit yardstick processed are converted, obtained the real-time diameter data of silicon rod;

4) image of processing in step 2 is transported in color comparison temperature measurement module 8, utilize the image processing system in the color comparison temperature measurement module image to be carried out to the correction of bad point, to Image Segmentation Using, level and smooth, layering, gradation conversion, then calling data storehouse, grey scale curve is carried out to matching, temperature is demarcated, completed thermometric;

5) by the silicon rod diameter and the surface temperature data that obtain in step 3,4, by data output interface, be transported in process optimization control module 3, in conjunction with work information, set up optimizing control models, with feedback, instruct silicon rod to be grown under the most rational processing condition;

6) the silicon rod diameter of above-mentioned acquisition and surface temperature data are outputed to terminal, obtain real-time diameter and the growth rate of silicon rod, and the real time temperature of silicon rod surface any point, and show in indicating meter 11.

Inventive principle of the present invention is as follows:

In the vision measurement module, according to the real-time growth diameter of the captured silicon rod growth image amount of measuring silicon rod.

(1) demarcation of camera system

The pixel value that will obtain in image is processed is converted into effective unit scale-value processed, needs calibration system, determines how much shooting models of camera.Known spatial target three-dimensional point homogeneous coordinates are

, the two-dimentional picture point homogeneous coordinates of its correspondence are

, spatial point

with picture point

between projective rejection be

$s\stackrel{?}{m}=A\left[\begin{array}{cc}R& t\end{array}\right]\stackrel{?}{M}$

Utilize least square solution overdetermination system of linear equations, provide external parameter; Solve inner parameter, if pick up camera, without lens distortion, can be solved by an overdetermination linear equation, if there is radial distortion, the Optimizing Search by a ternary solves.Finally solve inside and outside parameter, focal distance f, coefficient of radial distortion k, rotation matrix R and translation vector T.

(2) rim detection and Hough fitting of a straight line

Rim detection, according to the discontinuity of sensed luminance value, utilizes single order and second derivative to detect.The gradient of two-dimensional function f (x, y) is defined as vector

$\▿f=\left[\begin{array}{c}{G}_x\\ G_y\end{array}\right]=\left[\begin{array}{c}\∂f/\∂x\\ \∂f/\∂y\end{array}\right]$

This vectorial amplitude is

$\▿f=mag(\▿f)={[G_x^2+G_y^2]}^{1/2}={[{(\∂f/\∂x)}^2+{(\∂f/\∂y)}^2]}^{1/2}$

Generally be reduced to

$\▿f=G_x^2+G_y^2$ Or $\▿f=\left|G_x\right|+\left|G_y\right|$

Their values in constant brightness region are zero.

Use the Hough conversion to carry out the line detection and link at first doing the peak value detection, find and include peaked Hough converter unit, Hough converter unit in the maximum value neighborhood of a point found is made as to 0, repeats this step, until find the peak value needed.For each peak value, find the synthetic line segment of pixel groups that position is relevant.

(3) image corresponding point matching

Above-mentioned detected unique point is mated, found out the picture point in the different images of corresponding the same space point.

Suppose to take two width pictures for the silicon rod of synchronization, in two width images, the characteristic of correspondence point set is

, x (x, y, 1), x ' (x ', y ', 1) mean respectively the homogeneous coordinates of the corresponding points of left and right pick up camera.From the definition of fundamental matrix,

$\stackrel{\→}{{x^{\′}}^T}F\stackrel{\→}{x}=0$ ? $\stackrel{\→}{A}\stackrel{\→}{f}=0$

Wherein, $\stackrel{\→}{f}={(F_{11},F_{12},F_{13},F_{21},F_{22},F_{23},F_{31},F_{32},F_{33})}^T$ , meet constraint || f||=1;

$A=\left[\begin{array}{ccccccccc}{x_1}^{\′}{x}_1& {x_1}^{\′}{y}_1& {x_1}^{\′}& {y_1}^{\′}{x}_1& {y_1}^{\′}{y}_1& {y_1}^{\′}& x_1& y_1& 1\\ \·\·\·& \·\·\·\·& \·\·\·& \·\·\·& \·\·\·& \·\·\·\·& \·\·\·& \·\·\·& \·\·\·\\ x_n^{\′}{x}_n& x_n^{\′}{y}_n& x_n^{\′}& y_n^{\′}{x}_n& y_n^{\′}{y}_n& y_n^{\′}& x_n& y_n& 1\end{array}\right]$

Obtain fundamental matrix and accurately after Stereo matching point, under fundamental matrix instructs, carry out more corresponding point matching with the Robust estimation algorithm.

(4) foundation of optimizing control models

The process variable obtained by measurement, then set up optimizing control models in conjunction with work information, instruct the silicon rod growth rationally to carry out, and reduces costs, improves the quality of products to reach and the purpose of energy-saving and emission-reduction

The color comparison temperature measurement method of colorimetric temperature measurement system is as follows:

The general object that thermodynamic temperature is T, its radiation and distribution are described by the Planck radiation law:

$M(T,\mathrm{\λ})=\mathrm{\ϵ}\left(\mathrm{\λ}\right)c_1/{\mathrm{\λ}}^5[\exp \left(\frac{c_2}{\mathrm{\λT}}\right)-1]---\left(7\right)$

In formula, M _{e λ}for spectrum spoke out-degree, wherein λ is wavelength, and T is thermodynamic temperature, C ₁=3.741832 * 10 ^-12wcm ²for first radiation constant, C ₂=1.438786 * 10 ⁴μ mK is second radiation constant; Work as c ₂/ λ T>>1 o'clock, planck formula can be replaced by Wien equation, can be reduced to:

$M(T,\mathrm{\λ})=\mathrm{\ϵ}(\mathrm{\λ})c_1/{\mathrm{\λ}}^5\exp (\frac{c_2}{\mathrm{\λ}T})---\left(8\right)$

In λ T<2698 μ mK zones, the error of Wien equation and planck formula is less than 1%;

If two wavelength X ₁and λ ₂under, measure the spectrum spoke out-degree M(T sent by the object same point, λ simultaneously ₁) and M (T, λ ₂), temperature that can this point according to both ratio, formula is:

$T=\frac{c_2(\frac{1}{{\mathrm{\&lambda;}}_2}?\frac{1}{{\mathrm{\&lambda;}}_1})}{\ln \frac{M(T,{\mathrm{\&lambda;}}_1)}{M(T,{\mathrm{\&lambda;}}_2)}?\ln \frac{\mathrm{\&epsiv;}(T,{\mathrm{\&lambda;}}_1)}{\mathrm{\&epsiv;}(T,{\mathrm{\&lambda;}}_2)}?5\ln \frac{{\mathrm{\&lambda;}}_2}{{\mathrm{\&lambda;}}_1}}---\left(9\right)$

Suppose , approximately object is used as to grey body processes, formula (9) can be reduced to:

$T=\frac{c_2(\frac{1}{{\mathrm{\&lambda;}}_2}?\frac{1}{{\mathrm{\&lambda;}}_1})}{\ln \frac{M(T,{\mathrm{\&lambda;}}_1)}{M(T,{\mathrm{\&lambda;}}_2)}?5\ln \frac{{\mathrm{\&lambda;}}_2}{{\mathrm{\&lambda;}}_1}}----\left(10\right)$

This is the calculation formula of two-color thermometry.

Claims (3)

1. a polycrystalline silicon reducing furnace many reference amounts on-line monitoring and optimized control device, it is characterized in that, include many reference amounts infrared monitoring probe, infrared image processing and vision measurement module, process optimization control module, described many reference amounts infrared monitoring probe includes near-infrared optical system, image capture module, the water-cooled shield cap is equipped with in many reference amounts infrared monitoring probe outside, the front end of water-cooled shield cap is provided with water-cooling type and seals withstand voltage observation window, and many reference amounts infrared monitoring probe is deep in polycrystalline silicon reducing furnace, described infrared image processing and vision measurement module include infrared image processing module and vision measurement module, the infrared image processing module is connected data acquisition module successively, image processing module, the vision measurement module includes the color comparison temperature measurement module, the diameter measurement module, the output terminal of image processing module respectively with the color comparison temperature measurement module, the diameter measurement module connects, the color comparison temperature measurement module, the equal access data output interface of the output terminal of diameter measurement module, user interface, data output interface is circumscribed with the process optimization control module, the user interface external-connection displayer, the process optimization control module is connected with former Controlling System by a data interface, described process optimization control module is the closed optimized control model.

2. polycrystalline silicon reducing furnace many reference amounts on-line monitoring according to claim 1 and optimized control device, is characterized in that, described image capture module is the infrared CCD camera.

3. the method for a polycrystalline silicon reducing furnace many reference amounts on-line monitoring and optimal control, is characterized in that, comprises following concrete steps:

Many reference amounts infrared monitoring probe is by choosing optimum infrared band and highly sensitive infrared CCD camera, obtain the picture signal of silicon rod growth in stove by near-infrared optical system, infrared CCD camera, and carry out the A/D conversion through data acquisition module, obtain the infrared digital image of polysilicon rod;

2) infrared digital image of above-mentioned acquisition is sent into to image processing module, and carry out the image pre-treatment and cut apart through image processing module, in conjunction with concrete image, by Image denoising algorithm and level and smooth filtering algorithm, protected to greatest extent the imaging surface Characteristics of The Distribution of Temperature; Find out the edge of sealing by carrying out the profile tracking, and carry out sub-pixel rim detection and fitting of a straight line, obtain the image border of degree of precision;

3) image of above-mentioned processing is transported in the diameter measurement module of vision measurement module, at first the demarcation of carrying out the infrared CCD camera obtains the inside and outside parameter of camera, carry out choosing of primary objective image border according to the Hough Algorithm of fitting a straight line, then image is carried out the tracking at edge, zonule; Under instructing, the fundamental matrix of Robust estimation carries out corresponding point matching; Finally, according to above algorithm, the pixel value of measurement and effective unit yardstick processed are converted, obtained the real-time diameter data of silicon rod;

4) image of processing in step 2 is transported in the color comparison temperature measurement module, utilize the image processing system in the color comparison temperature measurement module image to be carried out to the correction of bad point, to Image Segmentation Using, level and smooth, layering, gradation conversion, then calling data storehouse, grey scale curve is carried out to matching, temperature is demarcated, completed thermometric;

5) by the silicon rod diameter and the surface temperature data that obtain in step 3,4, by data output interface, be transported in the process optimization control module, in conjunction with work information, set up optimizing control models, with feedback, instruct silicon rod to be grown under the most rational processing condition;

6) the silicon rod diameter of above-mentioned acquisition and surface temperature data are outputed to terminal, obtain real-time diameter and the growth rate of silicon rod, and the real time temperature of silicon rod surface any point, and show in indicating meter.

Images