JPH0849845A - Combustion deciding control method and deciding control device utilizing color image - Google Patents

Abstract

PURPOSE:To provide a combustion deciding method and deciding device which is devised to reduce a time delay to a value low enough to prevent a problem on a practicable use from arising. CONSTITUTION:A combustion state is photographed in a color image. A feature parameter determined from a chromaticity coordinate of a specified color specification widely related to a combustion state is inputted to a neutral network means 42 together with a process state amount, such as a combustion chamber outlet temperature T. From the image, adaptation to a typical combustion state is determined and the combustion state is evaluated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a combustion determination / control method and a determination / control device. More specifically, the present invention relates to a combustion determination / control method and a determination / control device using a color image.

[0002]

2. Description of the Related Art Conventionally, in a combustion device such as a refuse incinerator or a boiler, a combustion state is determined and a combustion diagnosis is performed in order to see whether or not the combustion is properly performed.
This is the case where combustion is not performed correctly, NO _X and CO and soot is generated, because cause air pollution. Further, combustion control is performed using the determination result and the diagnosis result to suppress the generation of NO _X , CO, and soot and dust.

Various proposals have heretofore been made regarding this combustion determination method and determination apparatus or combustion diagnosis method and diagnostic apparatus.

For example, Japanese Unexamined Patent Publication No. 1-263414 discloses a combustion diagnosis method for diagnosing a combustion state of a burner, in which a flame shape and a combustion process amount are measured, and a correlation between the flame shape and the combustion process amount is measured. A combustion diagnosis method based on flame shape measurement has been proposed, which is characterized by identifying a model representing the relationship and estimating the combustion process amount from the flame shape using the model.

However, in the combustion diagnosis method according to the proposal of Japanese Patent Laid-Open No. 1-263414, NO _X
Since it is necessary to measure the combustion process amount, which takes a long time to measure CO, O _2, and unburned matter in ash, combustion diagnosis is accompanied by a considerable time delay. Such a time delay does not become a serious problem when combustion requires several tens of minutes as in the case of a stoker combustion system, but when combustion is performed in a dozen seconds as in a fluidized bed furnace, for example,
The problem is that the diagnostic results do not match the actual combustion.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of the problems of the prior art, and provides a combustion determination method and a determination apparatus in which the time delay is reduced to such an extent that it does not pose a practical problem. The main purpose is to provide a combustion control method and a combustion control device using the result.

[0007]

Means for Solving the Problems As a result of intensive studies made by the present inventors on the problems of the prior art, not only the brightness of the flame but also the color change depending on the combustion state. It has been found that the present invention has a correlation with the above and that the correlation depends on the characteristic parameter obtained from the chromaticity coordinates of a specific color system.

That is, according to the first aspect of the present invention, the procedure for measuring the process state quantity, the procedure for capturing the combustion state with a color image, and the image processing of the obtained color image are performed for the red component, the green component and the blue component. A procedure of dividing the three-component image of the component into the first image memory and storing the same, and a procedure of converting the color system of the stored three-component image of the red component, the green component and the blue component,
A procedure for extracting a characteristic parameter greatly related to a combustion state from the image after the color system conversion, and using the extracted characteristic parameter and process state quantity as an input of a neural network, the compatibility with a typical combustion state And a method for determining combustion using a color image.

Here, the process state quantity is, for example, the combustion chamber outlet O ₂ concentration, the combustion chamber furnace pressure, or the combustion chamber outlet temperature.

In the first aspect of the present invention, it is preferable that a procedure for separating the flame region from the captured color image is added. Further, it is more preferable that a procedure for evaluating the combustion state by stochastically integrating the compatibility is added.

A second aspect of the present invention relates to a combustion control method using a color image, characterized in that combustion control is performed using a combustion determination or evaluation result by the combustion determination method.

More specifically, the combustion control adjusts the ratio of the air supply amount, the primary air amount and the secondary air amount, the water injection amount into the combustion chamber or the fuel supply amount.

A first aspect of the third aspect of the present invention comprises a process state quantity input processing means, a color image pickup means, a color system conversion means, a characteristic parameter extraction means and a neural network means, and the color system conversion is performed. The means converts the color system of the picked-up image, the characteristic parameter extraction unit extracts the characteristic parameter based on the component value of the converted color system, and the neural network unit processes the process state. The present invention relates to a combustion determination device using a color image, characterized in that the degree of compatibility with a typical combustion state is calculated and the combustion state is evaluated based on the process state quantity and the characteristic parameter from the quantity input means.

A second aspect of the third aspect of the present invention comprises a process state quantity input processing means, a color image capturing means, a color system conversion means, a characteristic parameter extraction means, a neural network means and a stochastic integration processing means. , The color system conversion means converts the color system of the captured image,
The characteristic parameter extraction means extracts characteristic parameters based on the converted component values of the color system, and the neural network means extracts typical characteristic values based on the process state quantity and the characteristic parameter from the process state quantity input means. Of the combustion state using a color image, in which the degree of compatibility with a specific combustion state is calculated, and the calculated degree of compatibility is stochastically processed by the stochastic integration processing means to evaluate the combustion state. Regarding the device.

Here, the process state quantity is the O ₂ concentration in the combustion chamber outlet, the combustion chamber furnace pressure, or the combustion chamber outlet temperature.

The first aspect of the fourth aspect of the present invention is to provide a process state quantity input processing means, a color image pickup means, a color system conversion means, a characteristic parameter extraction means, a neural network means, a combustion control processing means, and a process control amount. Output processing means, the color system conversion means converts the color system of the captured image, and the characteristic parameter extraction means generates characteristic parameters based on the converted component values of the color system. Extracted, the degree of compatibility with a typical combustion state is calculated based on the process state quantity and the characteristic parameter from the process state quantity input means, the combustion state is evaluated, and the combustion control process is performed. Means for calculating a process control amount based on the evaluation of the combustion state, and the process control amount output processing means for To the process control amount regarding combustion control apparatus using a color image, characterized in that the predetermined processing is outputted been made.

A second aspect of the fourth aspect of the present invention is to provide a process state quantity input processing means, a color image pickup means, a color system conversion means, a characteristic parameter extraction means, a neural network means, a stochastic integration processing means, and a combustion control. The colorimetric system conversion unit converts the colorimetric system of the captured image, and the characteristic parameter extraction unit converts the colorimetric component value. Based on the process state quantity and the characteristic parameter from the process state quantity input means, the degree of conformity with a typical combustion state is calculated by the neural network means, and the stochastic integration is performed. The processing means stochastically processes the calculated fitness to evaluate the combustion state, and the combustion control processing means performs the combustion control. Process control amount based on the evaluation of state is calculated by the process control amount output processing means,
The present invention relates to a combustion control device using a color image, wherein a predetermined process is performed on the process control amount and the process control amount is output.

Here, the process state quantity is, for example, a combustion chamber outlet O ₂ concentration, a combustion chamber furnace pressure, or a combustion chamber outlet temperature, and the process control quantity is, for example, an air supply amount, a primary air amount and It relates to the ratio of the amount of secondary air, the amount of water injected into the combustion chamber, or the amount of fuel supply.

[0019]

[Function] In addition to measuring the process state quantities such as the O ₂ concentration in the combustion chamber outlet, the pressure in the combustion chamber furnace, and the temperature in the combustion chamber outlet, a color image of the combustion state is taken, and red, green and blue are obtained from the obtained color image. A three-component image consisting of each color component is created, the color system conversion is performed on the three-component image, and the characteristic parameters greatly related to the combustion state are extracted from the image after the color system conversion. Input process state quantities and feature parameters into the neural network. In the neural network, the combustion state is determined or evaluated by obtaining the degree of compatibility with a typical combustion state based on the input process state quantity and characteristic parameter. However, since the color image of the combustion state is captured without time delay, the determination or evaluation of the combustion state using the image has no time delay in practical use.
Further, in a preferred aspect of the present invention in which the obtained fitness is stochastically integrated to evaluate the combustion state, stochastically integrated processing is performed, so that, for example, due to flame fluctuation or combustion unevenness, Even if the image changes irregularly, reliable judgment or evaluation can be made.

Further, if the combustion control is performed using the result of this judgment or evaluation, for example, the ratio of the air supply amount, the primary air amount and the secondary air amount, the water injection amount into the combustion chamber or the fuel supply amount is adjusted, As described above, the determination is made with no time delay, so that the combustion control can be performed by immediately responding to the combustion state. That is, optimal combustion control can be achieved.

[0021]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will now be described based on embodiments with reference to the accompanying drawings, but the present invention is not limited to such embodiments.

FIG. 1 shows a functional block diagram of a combustion determination / control apparatus used in the combustion determination / control method of the present invention. The determination / control apparatus A includes a color image pickup means 10 and a color system conversion means 22. , The characteristic parameter extracting means 24, the process state quantity input processing means 32, the neural network means 42, the stochastic integration processing means 44, the combustion control processing means 46, and the process control quantity output processing means 34 as main parts. Become.

The color image pick-up means 10 picks up an image of combustion and is, for example, a color television camera. The color image picked up by the color image pick-up means 10 is input to the color system conversion means 22 as a composite color image signal such as NTSC or an RGB three primary color component combination signal (hereinafter simply referred to as RGB signal) through a signal cable. It

The color system conversion means 22 converts the color system of the input signal. More specifically, the following processing is performed on the input color image signal. First, the input color image signal is stored in the RGB image memory (first image memory). When the input signal is a composite color image signal, it is converted into an RGB signal and then stored in this RGB image memory. Then RGB
The RGB color system image data stored in the image memory is converted to the color system, for example, the YIQ color system, and each component value, for example, (Y, I, Q) is calculated for each pixel. .

The characteristic parameter extracting means 24 extracts a characteristic parameter representing the combustion state of one screen from the obtained component values in the new color system. For example, an average value of I and Q per screen is calculated and the value is extracted as a characteristic parameter. The values of the extracted characteristic parameters are input to the neural network means 42 together with the process state quantity from the process state quantity input processing means 32. When extracting the characteristic parameter, it is preferable to perform a pre-processing for separating only the flame region so that the combustion state can be determined more accurately. The separation of the flame region is performed by, for example, performing threshold processing on the Y component value indicating the brightness. More specifically, an area having a Y component value equal to or larger than a certain value is set as a flame area, and an area having a Y component value less than the certain value is processed as another background area.

Here, the conversion from the RGB color system to the YIQ color system is performed by the following expression 1.

[0027]

[Equation 1]

The process state quantity input processing means 32 processes the signal state of the process state quantity such as the combustion chamber outlet O ₂ concentration, the combustion chamber furnace pressure, and the combustion chamber outlet temperature which are input as process signals from the combustion control device. It is performed and converted into a signal suitable for the processing of the combustion determination / control device A. The process state quantity input processing means 32 is, for example, an input interface.

The neural network means 42 obtains the degree of compatibility with a typical combustion state based on the values of the process state quantity and the characteristic parameter that have been input, for example, combustion deterioration, proper combustion, and active combustion. It makes a simple judgment or evaluation. Then, the judgment or evaluation result is input to the stochastic integration processing means 44. In addition,
If the stochastic integration processing means 44 is not provided,
The judgment or evaluation result is output as it is.

Here, the neural network means 42 is composed of an input layer, an intermediate layer and an output layer as shown in FIG. 2, and each layer is composed of a plurality of processing units. Each processing unit is combined with a weight coefficient determined by learning between the next layer and the previous layer, and each processing unit has an adder for summing the inputs from the previous layer and a threshold value determined by learning. It has a threshold function. In this learning of the neural network, an input pattern (teacher input) and a target output pattern (teacher output) pair is presented. Immediately after this presentation, the weights and thresholds are adjusted so that the difference between the network output and the target output is reduced. In learning, a learning set, which is a set of pairs of input patterns and target patterns, is used, and this is repeatedly presented to the network. When this learning is completed, the network operation test is performed.

More specifically, this learning includes a forward propagation step and a back propagation step executed thereafter. Both the forward propagation step and the back propagation step are executed each time a pattern is presented during learning. The forward propagation step begins with the presentation of the input pattern to the input layer of the network and continues as the activity level calculation proceeds forward through the intermediate layers.
All the processing units in each layer (shown by circles in FIG. 2) sum the inputs and calculate the outputs by a threshold function. The output layer of the unit then outputs the network.

Then, the output pattern of the network is compared with the target output pattern, and when there is a difference, the back propagation step is started. In the back-propagation step, the threshold value of the unit of each layer and the weight change amount are calculated, and this is traced backward in order from the output layer to the intermediate layer. In this backpropagation step, the network is adjusted for weights and thresholds so that the observed differences are reduced.

As a result of such learning, the output pattern from the network basically basically matches the target output pattern.

The stochastic integration processing means 44 cannot take an image of the entire combustion region because the image pickup by the color image pickup means 10 is performed through the viewing window. Therefore, for example, due to fluctuations in flames or uneven combustion. , Processing for preventing different judgments and evaluations from being made for images that are not too far apart. This processing will be described more specifically as follows.

Since the output pattern of the neural network means 42 is output as the degree of conformity with the respective patterns of deterioration of combustion, combustion appropriateness, and combustion activation, a certain combustion state has two or more patterns, for example, combustion deterioration, combustion appropriateness. In some cases, both patterns may be output when the matching degree is equal to or higher than a certain threshold. For such a state, the probabilistic integration processing means 44 calculates reliability intervals for a plurality of support signals by using Dempster &Shafer's probability theory, and quantifies the ambiguity between the conformances, and The evaluation result that is judged to be highly reliable is output.

The combustion control processing means 46 calculates a process control amount such as an air supply amount, a fuel supply amount, and a water injection amount into the combustion chamber according to the result of the evaluation of the combustion state.

The process control amount output processing means 34 is an air supply amount output as a process signal to the combustion control device,
The signal form of the process control amount such as the fuel supply amount and the water injection amount into the combustion chamber is processed and converted into a signal suitable for the process by the combustion control device. The process control amount output processing means 46 is, for example, an output interface.

The color system conversion means 22, the characteristic parameter extraction means 24, and the neural network means 4 are used.
2, stochastic integrated processing means 44 and combustion control processing means 4
6 is specifically realized by, for example, storing a program corresponding to the above functions in a board computer.

Hereinafter, the present invention will be described in more detail based on more specific examples.

Example FIG. 3 shows a schematic diagram of a combustion determination / control device used in this example. The determination / control device A processes a color image captured by the color television camera 10 connected via a cable to evaluate the combustion state and control the combustion.

The configuration of the color television camera 10 is not particularly limited, and various color television cameras conventionally used for capturing color images can be used.

The combustion determination / control device A is connected by a data bus and functions as a process state quantity input processing means 32 and a process control quantity output processing means 34, and a process signal input / output processing section 30 and color system conversion. The image processing unit 20 functions as the unit 22 and the characteristic parameter extracting unit 24, and the artificial intelligence processing unit 40 functions as the neural network unit 42, the stochastic integration processing unit 44, and the combustion control processing unit 46. for that reason,
The image processing unit 20 includes, for example, an image input board, a first image memory board (RGB image memory), a high-speed arithmetic board, a second image memory board, and an image processing CPU board.
The artificial intelligence processing unit 40 has, for example, a neural network, a stochastic integration processing unit, and a combustion control processing unit.
A development support computer 50 is connected to the artificial intelligence processing unit 40 of the combustion state determination / control device A for convenience of modifying the program in the test stage.

Thus, the color image from the color television camera 10 is converted into the chromaticity value of the RGB color system by the image input board, stored in the first image memory board, and then the RGB color system is displayed by the high speed operation board. The system is converted to the YIQ color system, and the component value thereof is calculated. Then, the characteristic parameter is extracted based on the component value. The extracted characteristic parameter is stored in the second image memory. It should be noted that here, the characteristic parameters are the respective component values of Y, I, and Q. Further, the value of the extracted characteristic parameter is input to the neural network of the artificial intelligence processing unit 40 by the image processing CPU board together with the process state quantity, for example, the combustion chamber outlet temperature.

The neural network evaluates deterioration of combustion, appropriateness of combustion, active combustion, etc. based on the value of the input characteristic parameter and the process state quantity (see FIG. 4). FIG. 5 shows an example of the evaluation result. As is apparent from FIG. 5, different evaluation results are obtained, that is, combustion deterioration and combustion activation. Therefore, Dempster & Sh
Probabilistic integration is performed using afer's probability theory. That is, the fitness when the combustion deteriorates, the fitness when the combustion is appropriate, and the fitness when the combustion is active are f _b , f _s , and f, respectively.
_a , and the integrated fitness values are f _bA , f _Ba ... f _bsA , respectively.
In other _words , the integrated fitness values f _xY and f _xyZ are _calculated by the following equations. _{f xY = (1-f x} ) · f y / (1-f x · f y) f xyZ = (1-f x) · f yZ / (1-f x · f yZ)

Where f _{bA is the} fitness for combustion activation from f _b , f _a f _{Ba is the} fitness for worsening combustion from f _b , f _{a ···} f _bsA : From f _b , f _s , f _a Combustion active fitness x, y, z: a, b, or c, which are different from each other. X, Y, Z: Any of A, B and C, which are different from each other.

The combustion state in the case of FIG. 5 is determined by the equation relating to the integrated fitness. From FIG. 5, f _b is 0.16
And f _a can be read as 0.94. Substituting these values into the above equation and calculating, the integrated fitness f _bA ,
f _Ba is 0.929 and 0.011, respectively. Therefore, it is evaluated that combustion is active.

Next, the correspondence between the evaluation result of the neural network and the characteristic change of the combustion chamber was investigated. The result is shown in FIG. From FIG. 6, an increase in the O ₂ concentration and CO concentration at the combustion chamber outlet is recognized after the signal indicating that combustion has deteriorated, and a decrease in the O ₂ concentration at the combustion chamber outlet after the signal indicating that combustion is active and An increase in CO concentration is observed. Therefore, it can be seen that the evaluation by the neural network matches the actual combustion. Also,
Since the combustion state is judged prior to the measured value of the characteristic change, it can be said that the combustion state is judged practically without time delay.

The learning of the neural network was performed using the development support computer 50 connected to the artificial intelligence processing unit 40. The details of the learning will be described below.

The neural network used here is
It consists of a hierarchical neural network consisting of an input layer, an intermediate layer and an output layer. Each cell in the middle layer is connected to all input layers. Each cell in the output layer is also connected to all cells in the intermediate layer. The output value O _i of each cell is represented by the Zigmoid function of the following equation. O _i = 1 / (1 + exp (-Σω _ij O _j ))

Here, ω _ij is a coupling coefficient from cell j to cell i, and its value is determined by learning using teaching data.

The back propagation method is used for learning, and when the teaching data T _i is given, the correction amount of ω _ij can be obtained by the following equation. Δω _ij (k) = αδ _j O _i + β Δω _ij (k−1) δ _j = (T _i −O _i ) O _i (1−O _i ).

Here, k represents the number of repetitions.

Further, as described above, the component values (characteristic parameters) of the outputs Y, I, Q of the image processing unit 20 and the combustion chamber outlet gas temperature T are input to the neural network for determining the combustion state. Was used.

Note that the output of the image processing unit 20 and the temporal change in the combustion chamber outlet gas temperature are also important information.
The value of each 1 second was measured at 30 points, which were also input to the neural network.

As a teaching data group for learning, time series data having the structure shown below is searched from the CO value and O ₂ value of the trend data after operation (see FIG. 6), and combustion deterioration is caused.
A total of 73 patterns were extracted for the three types of proper combustion and active combustion, and the teaching of the neural network was repeated 3000 times. As a result, the error for the obtained neural network teaching pattern was 0.59%.

Teaching data example (1) Neural network input data pattern NI ₁ : Y ₁₁ , Y ₁₂ ... Y ₁₃₀ , I ₁₁ , I ₁₂ ... I ₁₃₀ ,
_{Q 11, Q 12 ... Q 130} , T 11, T 12 ... T 130 · · · NI i: Y i1, Y i2 ... Y i30, I i1, I i2 ... I i30,
Q _i1 , Q _i2 ... Q _i30 , T _i1 , T _i2 ... T _i30 ... NI ₇₃ : Y ₇₃₁ , Y ₇₃₂ ... Y ₇₃₃₀ , I ₇₃₁ , I ₇₃₂ ...
I ₇₃₃₀ , Q ₇₃₁ , Q ₇₃₂ ... Q ₇₃₃₀ , T ₇₃₁ , T ₇₃₂ ... T ₇₃₃₀

(2) Neural network output data
Pattern NO ₁ : Deterioration of combustion ··· NO _i : Proper combustion · · · NO ₇₃ : Active combustion

Therefore, in the combustion control processing unit, in order to suppress the generation of CO, for example, with respect to the supply air amount, the fuel supply amount, and the water injection amount into the combustion chamber, according to the evaluation result of the combustion state. The following processing is performed.

(1) When it is determined that combustion deteriorates When combustion deteriorates, the reaction is frozen by quenching,
Further, since CO is generated by discharging the equilibrium composition gas at a high temperature, the following process is performed.

Primary air amount The primary air amount is increased. As a result, bed combustion is temporarily activated, and bed temperature and combustion chamber temperature are raised.

Secondary air amount The secondary air amount is reduced. Thereby, cooling of the combustion chamber is suppressed and the temperature of the combustion chamber rises.

Fuel supply amount The fuel supply amount is increased. As a result, deterioration of combustion due to fuel reduction is avoided.

Amount of Water Injection into Combustion Chamber The amount of water injection into the combustion chamber is increased. As a result, the water term in the Mizutani CO oxidation reaction formula shown below increases, the oxidation of CO is promoted, and the CO concentration decreases. -d [CO] /dt=1.2x10 ¹¹ [CO] [O ₂ ] ^0.3 [H ₂ O] ^0.5 exp (-
8050 / T)

[0064](2) When it is determined that combustion is active  When combustion becomes active, the mixing is uneven and local
CO is generated by exhausting fuel rich gas lumps
Therefore, the following processing can be performed.

Primary air amount Increase the primary air amount. This promotes gas mixing and eliminates O ₂ deficiency.

Secondary air amount The secondary air amount is increased. This promotes gas mixing in the combustion chamber and eliminates O ₂ deficiency.

Fuel supply amount The fuel supply amount is reduced. As a result, the amount of fuel is changed to an appropriate amount.

Amount of Water Injection into Combustion Chamber The amount of water injection into the combustion chamber is increased. As a result, the water term in the CO oxidation reaction formula increases, the oxidation of CO is promoted, and the CO concentration decreases.

Next, according to the above-mentioned procedure, for example, for evaluation of combustion deterioration or combustion activation of the stochastic integration processing unit,
It can be seen that when the secondary air amount is decreased or increased as shown in FIG. 6D, the generation of CO is suppressed as shown in FIG. 6B.

[0070]

As described above, in the present invention, the color image of the combustion state is image-processed together with the process state quantities such as the temperature of the combustion chamber outlet, the concentration of O _{2 of the} combustion chamber outlet, the pressure in the combustion chamber furnace and the like. The characteristic parameters are extracted by inputting them into a neural network to determine or evaluate the combustion state, so that the combustion state can be accurately determined or evaluated substantially without causing a time delay. That is an excellent effect.

In the preferred embodiment of the present invention, since the judgment or evaluation results are stochastically integrated to judge or evaluate the combustion state, the image is irregular in a short time due to flame fluctuation or combustion unevenness. Even if the fluctuations occur, it is possible to obtain an excellent effect that the combustion state can be accurately determined or evaluated without being affected by the fluctuations.

Further, in the combustion control according to the present invention, the process control amount such as the air supply amount, the ratio of the primary air and the secondary air, the water injection amount into the combustion chamber, the fuel supply amount, etc. is determined according to the judgment or evaluation. Since it is adjusted, it is possible to quickly respond to changes in the combustion state, and therefore, low-pollution combustion with suppression of CO and the like is possible, and an excellent effect of preventing air pollution is obtained.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a determination / control device used in a combustion state determination / control method of the present invention.

FIG. 2 is an explanatory diagram of a neural network used in the present invention.

FIG. 3 is a block diagram of a combustion state determination / control device according to an embodiment of the present invention.

FIG. 4 is an explanatory diagram of a neural network according to the embodiment.

FIG. 5 is a graph showing an example of evaluation results in the example.

FIG. 6 is a graph showing the relationship between the evaluation results and the combustion characteristics in Examples, where (a) is the output of the neural network, (b) is the CO concentration, and (c) is the O ₂ concentration,
The same (d) shows the amount of secondary air.

[Explanation of symbols]

10 Color Image Capturing Means, Color Television Camera 20 Image Processing Unit 22 Color System Converting Unit 24 Characteristic Parameter Extracting Unit 30 Process Signal Input / Output Processing Unit 32 Process State Quantity Input Processing Means 34 Process Control Quantity Output Processing Means 40 Artificial Intelligence Processing Unit 42 Neural Network Means 44 Stochastic Integration Processing Means 46 Combustion Control Processing Unit 50 Development Support Computer A Combustion Judgment / Control Device

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G05B 23/02 T 7531-3H 301 T 7531-3H G06F 15/18 550 E 8837-5L G06T 1 / 00 7/00 (72) Inventor Yasumitsu Kurosaki 1-1 Kawasakicho, Akashi-shi Kawasaki Heavy Industries Ltd. Akashi factory (72) Inventor Yuichi Miyamoto 1-1 Kawasaki-cho Akashi-shi Kawasaki Heavy Industries Ltd. Akashi factory (72) Kazuki Ogura 1-1 Kawasaki-cho, Akashi-shi, Kawasaki Heavy Industries Ltd. Akashi factory (72) Isao Tsukimoto 1-1 Kawasaki-cho, Akashi-shi Kawasaki Heavy Industries Ltd. Akashi factory (72) Inventor Hiroshi Fujiyama 1-3 1-3 Higashikawasaki-cho, Chuo-ku, Kobe-shi Kawasaki Heavy Industries, Ltd. Kobe headquarters (72) Inventor Norio Toyoshima 1-3-3 Higashi-kawasaki-cho, Chuo-ku, Kobe Kawasaki Heavy Industries, Ltd. Kobe Head Office (72) Inventor Hidetaka Miyazaki 1-1 Kawasaki-cho, Akashi-shi Kawasaki Heavy Industries Ltd. Akashi Factory (72) Inventor Yoshinobu Mori 1-1 Kawasaki-cho, Akashi-shi Kawasaki Heavy Industries Ltd. In-house Akashi Factory (72) Inventor Motohiro Inoue 1-1 Kawasaki-cho Akashi-shi Kawasaki Heavy Industries Stock Company In-house Akashi Factory

Claims (12)

[Claims] 1. A procedure of measuring a process state quantity, a procedure of capturing a combustion state by a color image, and a procedure of image-processing the obtained color image to divide it into a three-component image of a red component, a green component and a blue component. And a procedure for converting the color system of the obtained three-component image consisting of the red component, the green component and the blue component,
A procedure for extracting a characteristic parameter greatly related to a combustion state from the image after the color system conversion, and using the extracted characteristic parameter and process state quantity as an input of a neural network, the compatibility with a typical combustion state And a combustion determination method using a color image. 2. The process state quantity is the combustion chamber outlet O ₂
The combustion determination method using a color image according to claim 1, wherein the concentration, the combustion chamber internal pressure, or the combustion chamber outlet temperature is used. 3. The combustion determination method using a color image according to claim 1, wherein a procedure for separating a flame region from a captured color image is added. 4. The combustion determination method using a color image according to claim 1, wherein a procedure for evaluating the combustion state by stochastically integrating the compatibility is added. 5. A combustion control method using a color image, characterized in that combustion control is performed using a combustion determination or evaluation result by the combustion determination method according to claim 1. 6. The combustion control is performed by adjusting a ratio of an air supply amount, a primary air amount and a secondary air amount, a water injection amount into a combustion chamber or a fuel supply amount. A combustion control method using the color image according to item 5. 7. A process state quantity input processing means, a color image capturing means, a color system conversion means, a characteristic parameter extraction means, and a neural network means, wherein the color system conversion means expresses the color of the image captured. The system is converted, the characteristic parameter is extracted by the characteristic parameter extraction means based on the converted component value of the color system, and the process state quantity and the characteristic from the process state quantity input means are extracted by the neural network means. A combustion determination device using a color image, characterized in that the degree of compatibility with a typical combustion state is calculated based on the parameters to evaluate the combustion state. 8. A process state quantity input processing means, a color image capturing means, a color system conversion means, a characteristic parameter extraction means, a neural network means, and a stochastic integration processing means, wherein the color system conversion means captures an image. The converted color system of the image is converted, the characteristic parameters are extracted by the characteristic parameter extraction means based on the converted component values of the color system, and the neural network means outputs the process state quantity from the process state quantity input means. Based on the process state quantity and the characteristic parameter of, the compatibility with a typical combustion state is calculated, by the stochastic integration processing means,
A combustion determination device using a color image, wherein the calculated conformity is stochastically processed to evaluate a combustion state. 9. The process state quantity is a combustion chamber outlet O ₂
The combustion determination device using a color image according to claim 7 or 8, wherein the concentration, the internal pressure of the combustion chamber, or the outlet temperature of the combustion chamber. 10. A process state quantity input processing means, a color image pickup means, a color system conversion means, a characteristic parameter extraction means, a neural network means, a combustion control processing means, and a process control quantity output processing means The system conversion means converts the color system of the captured image,
The characteristic parameter extraction means extracts characteristic parameters based on the converted component values of the color system, and the neural network means extracts typical characteristic values based on the process state quantity and the characteristic parameter from the process state quantity input means. The degree of compatibility with a typical combustion state is calculated to evaluate the combustion state, the combustion control processing means calculates a process control amount based on the evaluation of the combustion state,
A combustion control device using a color image, wherein the process control amount output processing means performs a predetermined process on the process control amount and outputs the processed process control amount. 11. A process state quantity input processing means, a color image pickup means, a color system conversion means, a characteristic parameter extraction means, a neural network means, a stochastic integration processing means, a combustion control processing means, and a process control quantity output processing means. The colorimetric system conversion means, the colorimetric system of the captured image is converted, the characteristic parameter extraction means, the characteristic parameter is extracted based on the converted colorimetric component value, Based on the process state quantity and the characteristic parameter from the process state quantity input means by the neural network means, the compatibility with a typical combustion state is calculated, and by the stochastic integration processing means,
The calculated compatibility is stochastically processed to evaluate the combustion state, the combustion control processing means calculates a process control amount based on the evaluation of the combustion state, and the process control amount output processing means A combustion control device using a color image, wherein a predetermined process is performed on a process control amount and the process control amount is output. 12. The combustion chamber outlet O
₂ concentration, combustion chamber furnace pressure, or combustion chamber outlet temperature, wherein the process control amount relates to the ratio of the air supply amount, the primary air amount and the secondary air amount, the water injection amount or the fuel supply amount into the combustion chamber It is a thing, It is characterized by the above-mentioned.
Alternatively, a combustion control device using the color image described in 11 above.