JP2547190B2 - Method and apparatus for monitoring slag flow in partial combustion furnace of coal - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for monitoring downflow slag in a coal partial combustion furnace. More specifically, it relates to a slag downflow monitoring method and apparatus using a neural network in a partial combustion furnace for coal.

[0002]

2. Description of the Related Art Since coal has abundant reserves, active utilization of coal is being considered for stabilizing energy sources. However, it contains tens of percent of ash (alumina, silica, etc.) and harmful metals, so how to treat them is an important issue. In this regard, various proposals have hitherto been made such as a coal gasification method and a partial coal combustion method.

For example, regarding the coal gasification method, Japanese Patent Publication No.
-56079 discloses a coal which supplies coal and an oxidizer to a gasification chamber in a coal gasification furnace to gasify the coal and causes molten coal ash to flow down as slag into a slag cooling chamber below the gasification furnace. In the gasification method, the slag dropped from the gasification chamber to the slag cooling chamber is captured by an image, the slag drop amount and the drop frequency and the volume of the slag drop are measured by image processing, and the inside of the furnace is measured based on the measured value. A coal gasification method has been proposed which is characterized by diagnosing the suitability of temperature, coal supply amount and oxidant supply amount, and controlling coal supply amount and oxidant supply amount or deciding whether or not to stop operation.

Further, in Japanese Patent Laid-Open No. 64-24894, in a slag flow method of a coal gasifier in which coal ash is melted at a high temperature and discharged as slag, a slag to be dropped is captured by an image, Coal gasifier slag characterized by measuring the dropping frequency of slag and the volume of slag by image processing, calculating the amount of slag flowing down based on the measured value, and diagnosing the slag dropping state and the furnace state In the flow-down monitoring method, or in the slag flow-down method of the coal gasification furnace that melts coal ash under high temperature and discharges it as slag, an imaging device that captures the slag dripping situation, an image obtained by the imaging device An image processing device for processing and measuring the dropping frequency of the slag and the volume of the slag, and calculating the flow rate, and the slag dropping frequency, the volume and the flow rate obtained by the image processing apparatus. Dropping state, coal gasification slag flow-down monitoring device configured from a monitoring device for diagnosing the furnace state is proposed.

On the other hand, in the partial combustion method of coal, coal, air and oxygen are supplied at a high speed in the tangential direction of the cylindrical furnace, and a high-speed swirl flow is generated inside the combustion furnace to generate ash in the coal as slag and the bottom of the furnace. A partial combustion furnace for coal (hereinafter, referred to as CPC) for melting and removing is proposed (Kawasaki Heavy Industry Technical Report, No. 109, April 1991, pages 12 to 21). Even in this CPC, if the slag is not appropriately discharged, an accident such as slag accumulation in the furnace occurs,
It is necessary to monitor the flow condition of the slag and evaluate the flow condition.

However, in this CPC, since the slag is different from the above-mentioned coal gasification furnace in operating conditions and combustion system, it does not drop in a spherical shape as in the above-mentioned coal gasification furnace, but has an irregular shape. It has been confirmed that it has a shape and continuously flows down. Therefore, the above-mentioned Japanese Patent Laid-Open No. 64-24894
It is not possible to directly apply the slag monitoring method and evaluation method proposed in the Gazette. Therefore, since there is no suitable slag monitoring method or evaluation method at present, the slag flow-down state is visually observed by the operator. Then, judging from the experience of the operator, if it is recognized that the slag downflow port is blocked by the slag,
The lance is manually operated by the operator to remove the slag blocking the slag flow port. Alternatively, the lance is operated at regular intervals without monitoring the flowing state of the slag.

However, when the operator is required to monitor the slag flow-down state, the operator must constantly monitor the slag flow-down state, so that there is a problem that the operator is forced to perform long-time simple labor. Also,
Since the work is a simple work, there is a risk that the monitoring will be neglected due to looseness of the air, leading to a blockage of the outlet.
Further, there is a problem that it is difficult to secure human resources because it requires skill to judge the flowing state of the slag and the fact that such simple work is disliked.

On the other hand, when the lance is operated at regular intervals regardless of the slag flow-down state, the lance work may be performed even when the slag flow-out port is not closed, resulting in excessive lance work. There is a problem that the lance is activated and the lance is deteriorated and the economical efficiency of the lance is impaired. There is also a problem that unnecessary lance work may cause slag clogging.

[0009]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems of the prior art, and it is a main object of the present invention to provide a slag monitoring method and apparatus capable of automatically evaluating the flowing state of slag. Has an aim.

[0010]

Means for Solving the Problems The inventors of the present invention have conducted extensive studies to solve the problems of the prior art, and as a result, the slag flow-down state is classified into several typical types from a stable state to an unstable state requiring attention. The slag can be classified into various patterns, and the degree of conformity with these typical patterns can be obtained by a neural network using the error backpropagation method from the image of the flowing state of the slag, the slag temperature, the furnace pressure, etc. The inventors have found a method for evaluating the flow-down state and completed the present invention.

That is, the slag flow monitoring method for the coal partial combustion furnace of the present invention comprises a procedure for obtaining an image of the slag flowing down and an area of the slag flow portion by the image processing of the image.
A procedure for measuring the luminance distribution and shape, a procedure for obtaining a goodness of fit with a typical slag flow-down state using a neural network with the measured values as an input, and a slug flow-down by stochastically integrating the goodness of fit. And a procedure for assessing the condition.

In the slag flow monitoring method for the partial combustion furnace of the present invention, the centroid dispersion of the slag within a certain time is also measured in the image processing of the image, and the measured value is added to the input of the neural network. Is preferred.

Further, in the slag flow monitoring method for the partial combustion furnace of the present invention, it is preferable to measure the temperature of the flowing slag and add the measured value to the input of the neural network.

The slag flow monitoring device for a partial combustion coal furnace of the present invention comprises an image pickup means, an image processing means, a neural network means, and a stochastic integration processing means,
An image of the slag flowing down is captured by the image capturing means, the area of the slag flow portion, the luminance distribution and the shape are measured by the image processing means, and the neural network means receives the measured value as a typical slug flow-down state. Of the slag is evaluated by stochastically processing the suitability by the stochastic integration processing means.

In the apparatus for monitoring the slag flow of the partial combustion furnace for coal according to the present invention, the dispersion of the center of gravity of the slag within a fixed time is also measured by the image processing means, and the measured value is added to the input of the neural network. Is preferred.

Further, in the slag flow monitoring device for the partial combustion furnace of the present invention, temperature detecting means is added, the temperature of the slag flowing down is measured by the temperature detecting means, and the measured value is measured by the neural network. Is preferably added to the input of.

[0017]

According to the present invention, the area, brightness distribution and shape of the slug flow are measured from the image of the slag flowing down, and the neural network is used as the input of the measured values to find the compatibility with the typical slug flow-down state. Then, the conformity is stochastically processed to evaluate the slag flow-down state. Therefore, it is possible to accurately grasp the flow of the slag flowing in an indeterminate and continuous manner. Further, since a so-called back-propagation neural network is used, its learning function improves the accuracy of determination of the fitness as the driving time elapses. Therefore, after some time has passed,
Even if the operator does not monitor the flowing state of the slag, it can be operated safely.

Further, in the preferred embodiment of the present invention in which the center of gravity distribution of the slag flow is also obtained and used as the input of the neural network, the compatibility with the typical slug flow-down state can be obtained more accurately.

Further, the temperature of the flowing slag is measured,
In another preferred embodiment of the invention in which the measurements are added to the input of the neural network, the operating conditions of the furnace will also be taken into account in the determination of the slag flow-down condition,
The degree of compatibility with the typical slag flow-down state can be obtained even more accurately.

[0020]

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.

Embodiment 1 FIG. 1 shows a functional block diagram of a monitoring device used in a CPC slag downflow monitoring method according to Embodiment 1 of the present invention. The device K includes an image pickup means 10, an image processing means 20, and a neural network. The means 30 and the stochastic integration processing means 40 are the main constituent elements.

The image pickup means 10 is for picking up an image of the slag which is set near the slag outlet and flows down.
As the image pickup means 10, for example, a CCD camera is used. However, the present invention is not limited to this, and an infrared camera or the like can be preferably used.

The image processing means 20 extracts the slag portion from the image input from the image pickup means 10, and the area, luminance distribution, shape (width ratio, width average, etc.) of the slag flow portion from the extracted image and It is to measure the centroid dispersion. As the image processing means 10 having such a function, for example, an image processing device including an image processing processor and a computer is used.

Here, the extraction of the slag portion from the image of the slag flow-down state is performed by the binarization process utilizing the recognition by the luminance, but since this image contains a lot of noise, it is simply 2 There are cases where only the slag part cannot be extracted only by performing the binarization process. At this time, noise having a relatively small size is removed by performing contraction / expansion processing. As for the noise due to the combustion flame in the combustion furnace that leaks from the slag downflow port, the time from the occurrence to the disappearance is shorter than the change in the slag downflow state, so a fixed time, for example, 0.3 seconds, is used for imaging. An AND process is performed on the generated two slag flow-down images to remove the images.

Then, when the brightness of the image is generally high and there is not much difference in brightness between the slag portion and other portions, histogram smoothing processing is performed to increase the contrast between the slag portion and the background, and then the binary value is set. The slag portion is extracted by subjecting to the slag process.

Then, from the image of the slag portion thus obtained, the area of the slag portion, the brightness distribution width, the shape, and the centroid dispersion within a fixed time are measured. Here, the brightness distribution width is defined by the difference between the maximum value and the minimum value of the brightness of the slag portion extracted from the image in the slag downflow state, and the shape is a constant interval with the slag downflow port as a reference. Defined by the width of the slag part (eg, upper, middle, lower 3 places) and the ratio based on a certain part of the slag (eg, upper is 1, middle / upper, lower / upper) What is done. Further, the constant time when measuring the centroid dispersion is, for example, 6 seconds at intervals of 2 seconds.

The neural network means 30 has a function of performing image recognition from the characteristic parameter group,
For example, as shown in FIG. 2, it has a three-layer structure. Here, the error back propagation used in this neural network will be briefly described.

Input patterns and target output pattern pairs are presented to the network. Immediately after each presentation, the weights are adjusted so that the difference between the network output and the target output decreases. A training set, which is a set of pairs of input patterns and target patterns, is used for training, and this is repeatedly presented to the network. Test the network behavior after training.

The back propagation learning algorithm includes a forward propagation step and a back propagation step executed thereafter. Both the forward propagation step and the backward propagation step are executed each time the pattern is presented during training. The forward propagation step starts with the presentation of the input pattern to the input layer of the network (corresponding to the first layer in FIG. 2) and the calculation of the activity level through the hidden layer (corresponding to the second layer in FIG. 2). Continues while forward propagating. All the processing units in each layer (indicated 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.

The output pattern of the network is compared with the target vector, and when there is a difference or an error in the target vector, the back propagation step is started. In the back-propagation step, the error value of the unit in the hidden layer and the change amount of the weight are calculated. Starting from the output layer, this is followed in order in the hidden layer in the opposite direction. In this backpropagation step, the network corrects the weights to reduce the observed errors.

In the present embodiment, there are four types of teaching data from a normal state to a state where stable extraction of slag may be difficult (1: abnormal, 2: substantially abnormal, 3: substantially normal, 4: normal). The slag flowing state of was used. Then, in the learning, the learning is repeated until the mean square error with the teaching data in the output layer becomes 5% or less.

FIGS. 3 and 4 show the compatibility with a typical slug flow-down state by such a neural network. In the case of using this neural network, as is clear from FIG. 3, the adaptability to different states may simultaneously become high output values. Therefore, in this embodiment, the processing by the stochastic integration processing means described later is performed.

Although the center-of-gravity dispersion within a fixed time is also used as a characteristic parameter here, this center-of-gravity distribution may be used as necessary and is not necessarily used. Further, as the characteristic parameters, slag temperature, furnace pressure, etc. can also be used (see FIG. 8).

The neural network means 30 having such a function can be constructed by storing a program for performing the above-mentioned error back propagation processing in a normal engineering workstation.

The probabilistic integration processing means 40 probabilistically processes the goodnesses of fit for different states as described above in order to make an appropriate judgment in case the highness values are output at the same time. Is. The probabilistic integrated processing means 40 is also realized by storing a program for performing the following processing in the engineering workstation.

Next, regarding this probabilistic integration processing, FIG.
This will be described with reference to FIGS.

FIG. 3 is a graph showing the result of the goodness of fit by the neural network at a certain time, and FIG. 4 is 4 seconds after that. Here, FIG.
Since the type 4 has the highest degree of conformity, the degree of conformity of other types is corrected by setting this to 1 and the result is shown in FIG. On the other hand, in Fig. 4, type 1 has the highest compatibility, so
The same process is performed with this set to 1, and the result is shown in FIG. After that, these were synthesized and the results are shown in FIG. 7.

Next, the composite basic probability is obtained based on this composite result. In this case, conflicting combinations such as the combination of type 4 and type 1 are deleted in advance and normalized so that the total sum becomes 1. T (4) = (+) / (1-) = (0.23x0.11 + 0.23x0.54) / (1-0.23x0.35) = 0.163 T (1) = (+) / (1-) = (0.35x0.20 + 0.35x0.57) / (1-0.23x0.35) = 0.293 T (4,1) = (++) / (1-) = (0.11x0.20 + 0.11x0.57 + 0.54 x0.57) / (1-0.23x
0.35) = 0.427 T (4,1,2) = / (1-) = 0.54x0.20 /(1-0.23x0.35) = 0.117 where T (4): Type 4 is limited to a solution Probability that can be T (1): Probability that can be limited to type 1 as a solution T (4,1): Probability that can be said to be either type 4 or type 1 T (4,1,2): Type 4 Probability of not being either type 1 or type 2

Therefore, the possibility that each type will be a solution is as follows. Type 1: 0.837 Type 2: 0.117 Type 3: 0.0 Type 4: 0.707

From the above, the confidence intervals of type 1 and type 4 are 0.293 to 0.837 and 0.163 to 0.707, respectively. From this, it can be determined that the current slag flow-down state is type 1 and is in an abnormal state.

As described above, in this embodiment, even if different output results are obtained, they can be integrated and evaluated.

Embodiment 2 FIG. 8 shows a functional block diagram of a monitoring device used in the CPC slag downflow monitoring method according to the second embodiment of the present invention. The detection means 50 and the in-furnace pressure detection means 60 are added. Only the points of the second embodiment different from the first embodiment will be described below.

As the temperature detecting means 50, for example, a temperature detecting device including a radiation thermometer and an amplifier is used. As the in-furnace pressure detecting means 60, for example, a pressure detecting device including a pressure sensor and an amplifier is used.

The measurement results of the temperature detecting means 50 and the furnace pressure detecting means 60 are input to the neural network means 30 together with other measurement results. And Example 1
By performing the same processing as the above, the compatibility with the typical slag flow-down state is calculated. In Example 2, since the temperature of the slag flow and the pressure in the furnace are also used as inputs to the neural network, it is possible to improve the reliability of the degree of conformity with the typical pattern of the slag flow as compared with Example 1. it can.

The adaptability thus obtained is subjected to the probabilistic integration process in the same manner as in the first embodiment to determine whether the operation is abnormal.

As described above, in the second embodiment, since the physical quantities closely related to the state of the slag flow such as the pressure in the furnace and the temperature of the slag flow are taken into consideration, more reliable monitoring can be performed.

Embodiment 3 FIG. 9 shows a functional block diagram of an apparatus used in a CPC control method according to Embodiment 3 of the present invention. In the third embodiment, the signal from the first embodiment is input to the control means 70 to control the CP.
The amount of air supplied to C, the amount of coal supplied, or both are controlled.

The control of the CPC by the control means 70 is as follows.
For example, it is done as follows.

A typical slug flow can be associated with a physical quantity as follows, for example.

Therefore, in consideration of the above relationship, for example, the relationship between the slag clogging and the amount of supplied coal, the amount of supplied air, the slag viscosity, and the amount of slag is obtained in advance, and a normal slag flow is obtained from the input slag flow state. The amount of operation required for shifting to the flowing state, for example, the amount of supplied coal and the amount of supplied air are calculated, and, for example, the coal supply device and the blower are adjusted so as to match the operated amounts. In this case, when it is input that the flowing state of the slag is normal, the conventional control is continued.

Although a detailed description of the configuration of the control means 70 is omitted, the signal input from the stochastic integration processing means 40 can be processed together with the signals from other control elements. Other than that, it is configured similarly to the control device conventionally used for this type of control.

Embodiment 4 FIG. 10 shows a functional block diagram of an apparatus used in a CPC control method according to Embodiment 4 of the present invention. In the fourth embodiment, the signal from the second embodiment is input to the control means 70 and C
The amount of air supplied to the PC, the amount of coal supplied, or both are controlled.

The control of the CPC by the control means 70 is performed in the same manner as in the third embodiment. Also, this control means 70
Also, it is configured similarly to the control means of the third embodiment.

Embodiment 5 FIG. 11 shows a functional block diagram of an apparatus used in a lance control method according to Embodiment 5 of the present invention. In this fifth embodiment, the signal from the first embodiment is input to the control means 70 to control the operation of the lance 80.

The control of the lance by the control means 70 is as follows.
For example, it is done as follows.

When it is input that the flowing state of the slag is abnormal, the lance is operated to perform lancing, and the clog of the slag is removed.

When an input indicating that the slag flow-down state is substantially abnormal is made, the time change of the input is monitored, and when the input indicating that the slag flow-down state is substantially abnormal continues for a certain time, the lance is turned on. Activate to perform lancing. This fixed time is appropriately selected according to the CPC applied.

When an input is made that the slag flow-down condition is almost normal, the lance is not activated and no lancing is performed.

When it is input that the slag is flowing down normally, the lance is not activated and the lancing is not performed in this case as well.

Although the detailed description of the structure of the control means 70 is omitted, the signal input from the stochastic integration processing means 40 can be processed together with the signals from other control elements. Other than that, it is configured similarly to the control device conventionally used for this type of control.

Sixth Embodiment FIG. 12 shows a functional block diagram of an apparatus used for a CPC control method according to a sixth embodiment of the present invention. In the sixth embodiment, the signal from the second embodiment is input to the control means 70 to control the lance 80.

The control of the lance by the control means 70 is as follows.
The same procedure as in Example 5 is performed. Further, this control means is also configured similarly to the control means 70 of the third embodiment.

Next, the structure of the slag flow-out port obstruction prevention device for performing lancing will be described with reference to FIGS.

The slag adhering to and flowing down on the inner peripheral wall of the CPC enters the slag cooling layer arranged at the lower part through the slag flow-out port, is cooled and solidified, and is then carried out by a conveyor or the like. . At this time, in the slag flow-down opening part, the slag is often solidified due to the heat difference between the upper combustion furnace and the lower cooling tank, and the slag flow-down opening part adheres and grows to block the slag flow-down opening part. 13 to 1
Reference numeral 5 shows a device for preventing such blockage.

First, in the apparatus 80A shown in FIG. 13, a plurality of oxygen blow-out pipes 81 are arranged according to the width of the slag flow-down port 91 in the portion of the slag flow-down port 91 below the combustion chamber furnace wall, and the combustion furnace 90
Oxygen is ejected from the oxygen outlet pipe 81 during combustion inside. The slag downflow port 91 is a combustion furnace 90.
Since it is in a high temperature atmosphere near the inside, the ejected oxygen reacts with the combustible gas or char in the combustion gas at the slag downflow port 91 and burns, and also oxidizes with the residual carbon in the molten slag. By the slag outlet 91
The ambient temperature of the part is raised, and the slag that is being cooled and solidified is melted again and allowed to flow down into the cooling tank disposed below. This can prevent the slag downflow port 91 from being blocked by the slag. Here, since the oxygen ejection pipe 81 is a fixed type, there is a possibility that it will be burnt out by radiant heat from the inside of the combustion furnace 90 or combustion heat at the time of reaction between ejected oxygen and combustible material, and therefore the tip portion is cooled. By being stored in the water box 82, it is prevented from being burnt out.

A device 80B shown in FIG. 14 is a burner capable of supplying combustion air and fuel combustion air to a movable water-cooled slag cutter 83 which can be freely inserted / removed at the position of the oxygen blowing pipe 81 in the device 80A shown in FIG. 84 is provided. When removing the icicle-like slag generated in the slag downflow port 91, the burner 84 is used when the icicle-like slag is in a thin state without sufficiently growing, or in a state before being sufficiently solidified. It is removed only by the slag cutter 83 without burning. On the other hand, when the slag grows and the thickness increases, or when the cooling and solidification progresses and the strength of the slag increases, the operation of only the slag cutter 83 may damage the surrounding refractory material during the slag removal. . In that case, a fuel such as LPG and the air necessary for burning this fuel are fed from a burner 84 provided in the slag cutter 83, and the slag solidified by burning is made to flow again so that the slag flows down. Alternatively, while the slag is returned to the molten state by heating the burner 84, the slag cutter 83 is operated to cut the slag, and the slag is prevented from being blocked by dropping the slag into the cooling tank disposed in the lower part. obtain.

The apparatus 80C shown in FIG. 15 is replaced with the burner 84 in the apparatus 80B shown in FIG.
5 is provided in a movable water-cooled slag cutter 83 that can be freely inserted and removed. Thereby, even if the slag cannot be removed by the operation of the slag cutter 83 alone, when the ambient temperature near the slag is high, the oxygen ejected from the ejection pipe 85 reacts with the combustible gas or char in the combustion gas to burn. In addition, the residual carbon or the like in the molten slag is also oxidized to raise the atmospheric temperature of the flow-down port 91, so that the slag that is being cooled and solidified is melted again to flow down, or the cutting by the slag cutter is facilitated.

Although the present invention has been described above by taking the case where it is applied to the slag of the partial combustion furnace for coal as an example, the present invention is not limited to such an example, and slag flowing down in an indeterminate form is generated. It can be preferably used for a combustion furnace.

[0069]

As described above, according to the present invention,
Neural network is used for the image of slug flow to calculate the goodness of fit with typical downflow pattern of slug flow, and it is probabilistically integrated and evaluated. Can be evaluated accurately and automatically. Therefore, it is possible to obtain the effect that the operator does not need to monitor the flowing state of the slag. Further, in the present invention, since the slag flow-down state is evaluated by a neural network, the learning function of the slag improves the accuracy of the evaluation with the accumulation of operating time, and the operating conditions such as changes in coal type Can easily respond to changes.

Furthermore, if the lance is controlled based on the monitoring result obtained by the slag flow monitoring method of the present invention, the result that the lance can be performed efficiently can be obtained. with this,
There is an effect that unnecessary wear of the lance is prevented and its economical efficiency is improved.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a slag downflow monitoring device according to a first embodiment of the present invention.

FIG. 2 is a schematic diagram of a neural network in the same embodiment.

FIG. 3 is a graph showing a degree of compatibility with a typical slag flowing state in the example.

FIG. 4 is a graph showing the compatibility with a typical slag flow-down state in the example, which is 4 seconds after the state of FIG. 3;

5 is a graph obtained by modifying the graph of FIG. 3 with type 4 set to 1. FIG.

FIG. 6 is a graph obtained by modifying the graph of FIG. 4 by setting type 1 to 1.

7 is a composite of the graph of FIG. 5 and the graph of FIG.

FIG. 8 is a functional block diagram of a slag downflow monitoring device according to a second embodiment of the present invention.

FIG. 9 is a functional block diagram of a CPC control device using the slag downflow monitoring device according to the first embodiment of the present invention.

FIG. 10 is a functional block diagram of a CPC control device using a slag downflow monitoring device according to a second embodiment of the present invention.

FIG. 11 is a functional block diagram of a lance control device using the slag downflow monitoring device according to the first embodiment of the present invention.

FIG. 12 is a functional block diagram of a lance control device using the slag downflow monitoring device according to the second embodiment of the present invention.

FIG. 13 is a schematic view of an example of a slag downflow prevention device.

FIG. 14 is a schematic view of another example of the slag downflow blockage preventing device.

FIG. 15 is a schematic view of another example of the slag downflow blockage preventing device.

[Explanation of symbols]

 10 Imaging Means 20 Image Processing Means 30 Neural Network Means 40 Probabilistic Integrated Processing Means 50 Temperature Detecting Means 60 Reactor Pressure Detecting Means 70 Control Means 80 Lances 90 CPC (Combustion Furnace) K Slag Downflow Monitoring Device

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Masami Kobayashi 1-1 Kawasaki-cho, Akashi-shi Kawasaki Heavy Industries Ltd. Akashi Plant (72) Kenichi Fujii 1-1 Kawasaki-cho Akashi-shi Kawasaki Heavy Industries Ltd. Akashi Inside the factory (72) Inventor Eiichi Harada 1-1 Kawasaki-cho, Akashi City Kawasaki Heavy Industries Ltd. Akashi Factory (72) Inventor Shuichiro Ozeki 1-1 Kawasaki-cho Akashi City Kawasaki Heavy Industries Ltd. Akashi Factory (56) References JP-A-64-24893 (JP, A)

Claims (11)

(57) [Claims] 1. A procedure for obtaining an image of slag flowing down continuously, a procedure for measuring an area, a luminance distribution and a shape of a slag flow portion by image processing of the image, and a neural network which receives the measured value as an input. Calculating the goodness of fit by using a typical slag flow condition
A slag flow monitoring method for a partial combustion furnace of a coal, comprising: a discharge procedure; and a stochastic integrated processing procedure for stochastically integrating the suitability to evaluate a slag flow state. 2. A procedure for obtaining an image of slag flowing down, By the image processing of the image, the area and brightness of the slag flow part
A procedure for measuring distribution and shape, Using a neural network with the measured values as input
Of a goodness-of-fit calculation to find the goodness of fit with typical slag flow conditions
Out procedure, The suitability is stochastically integrated to determine the flow state of the slag.
Partial combustion of coal comprising stochastic integrated processing procedure to evaluate
A method for monitoring the slag flow of a furnace, By the stochastic integration processing procedure,
The calculated fitness at a certain time and the predetermined time
Later, it differs from the fitness calculated by the above-mentioned fitness calculation procedure.
If so, calculate the combined basic probability by combining both goodness of fit,
Evaluate the slag flow-down state based on the calculated combined basic probability
Monitoring of slag flow in a partial combustion furnace for coal
Method. Wherein also measuring the centroid variance within a fixed time of the slag in the image processing of the image, the coal according to claim 1 or 2, wherein the adding a measured value to the input of the neural network Monitoring method for slag flow in partial combustion furnace. 4. The method for monitoring the slag flow in a partial combustion furnace for coal according to claim 1 , 2 or 3 , wherein the temperature of the slag flowing down is measured and the measured value is added to the input of the neural network. 5. The operation of the lance is controlled according to the evaluation of the slag flow-down state obtained by the slag flow-down monitoring method for a partial combustion coal furnace according to claim 1, 2 , 3, or 4. Lance control method for partial combustion furnace of coal. 6. A coal supply amount and / or an air supply amount are controlled in accordance with an evaluation of a slag flow-down state obtained by the slag flow-down monitoring method for a coal partial combustion furnace according to claim 1, 2 , 3 or 4. A method for controlling a partial combustion furnace for coal, comprising: 7. An image pickup means, an image processing means, a neural network means, and a stochastic integration processing means are provided, wherein an image of the slag flowing down is taken by the image pickup means, and the slag flow is taken by the image processing means. The area, the brightness distribution and the shape of the part are measured, the degree of conformity with a typical slag flow-down state is obtained by inputting the measured value by the neural network means, and the degree of conformity is stochastically calculated by the stochastic integration processing means. A slag flow monitoring device for a partial combustion furnace of a coal, which is characterized in that the slag flow condition is evaluated by treating the slag. 8. The coal partial combustion furnace according to claim 7 , wherein the center of gravity distribution of the slag within a certain time is also measured by the image processing means, and the measured value is added to the input of the neural network. Slag downflow monitoring device. 9. temperature detecting means becomes addition is specified, a temperature of the slag flowing down the temperature detection means is determined, according to claim 7 or 8, characterized in adding a measured value to the input of the neural network A slag downflow monitoring device for the partial coal combustion furnace described. 10. A slag downflow port obstruction prevention device, which is controlled by an evaluation result signal input from the slag downflow monitoring device of the partial coal combustion furnace according to claim 7, 8, or 9 . 11. A coal characterized in that the coal supply amount and / or the air supply amount is controlled by an evaluation result signal input from the slag flow monitoring device of the partial coal combustion furnace according to claim 7, 8 or 9. Partial combustion furnace.