JP6983684B2 - Control device, boiler, boiler monitoring image acquisition method and boiler monitoring image acquisition program - Google Patents

Description

The present invention relates to a control device, a boiler, a method for acquiring a surveillance image of a boiler, and a program for acquiring a surveillance image of a boiler.

A pulverized coal-fired boiler is known in which pulverized coal obtained by crushing coal is used as pulverized fuel (solid fuel), the pulverized coal is burned by a combustion burner, and the heat generated by the combustion is recovered.
Deposits such as ash (clinker) due to combustion adhere to the furnace wall, burner, etc. of this pulverized coal-fired boiler, and when this deposits, the burner nozzle is blocked, the heat transfer characteristics of the panel deteriorate, and the deposited deposits. May fall and damage the furnace. Therefore, it is required to monitor the adhesion status of deposits.

However, when the pulverized coal-cooking boiler is in operation, the radiation of the pulverized coal flame is strong, so the operator can visually check the adhesion status of the deposits on the furnace wall such as the burner and panel in the furnace or with a visible light camera. It's difficult to do. Therefore, when observing the situation inside the furnace, it is necessary to reduce the load on the boiler and switch to an oil-fired burner, or stop the plant.

Further, as for the observation of the adhesion state of the deposits, since the operator visually observes and confirms the image with the visible light camera, the accuracy is greatly affected by the skill level of the operator.

Therefore, in order to observe the situation inside the furnace while continuing the operation, it is considered to use an infrared camera instead of a visual or visible light camera.

For example, Patent Document 1 discloses that a thermal image is provided by using an infrared image pickup device for measuring the temperature of a tube in a furnace. Further, Patent Documents 2 and 3 disclose that the inside of the ash melting furnace is monitored by using an infrared camera.

In addition, in order to save labor in observing the state of adhesion of deposits by workers, it is being considered to perform machine learning based on captured images and utilize the learning results.
In recent years, image recognition technology using machine learning has been rapidly developed, and for example, a technology capable of discriminating an object from an image with an accuracy close to that of a human is being put into practical use. In particular, the machine learning method called Deep Learning has been remarkably developed, and is expected to be put into practical use in the field of image recognition in combination with the improvement of computational function.

For example, in Patent Document 4, the monitoring device for the molten slag flow measures the three-dimensional shape of the slag, and the quality of the slag discharge property is determined by pattern recognition based on past cases, and the quality of the slag discharge property is determined. It is disclosed that the characteristics of the reactor are learned by adding the input information determined by the above to the database as new case information. Further, Patent Document 5 discloses that the slag flow state is evaluated by obtaining the conformity with a typical pattern from an image of the slag flow state and the slag temperature by using a neural network.

Special Table 2006-518464 Japanese Patent No. 4917950 Japanese Patent No. 5261038 Japanese Unexamined Patent Publication No. 2006-118744 Japanese Unexamined Patent Publication No. 7-126663

However, in the invention disclosed in Patent Document 1, since the thermal image is acquired, there is a problem that the adhesion state of the deposit cannot be confirmed. Further, in the inventions disclosed in Patent Documents 2 and 3, the floating matter of the ash melting furnace is transmitted by the infrared camera, but the observation range is due to the combustion state in the furnace and the dust caused by the fluctuation of the pulverized coal concentration. There was a problem that the monitored part could be observed only for a moment because it was blocked. As described above, since the quality of the captured image is not good, even if the image recognition technique using this image as an input is applied, there is a possibility that the desired result cannot be obtained.

Further, in the inventions disclosed in Patent Documents 4 and 5, the state is evaluated by adapting the shape of the slag to the pattern, and there is a problem that machine learning is used only for determining the state. there were.

The present invention has been made in view of such circumstances, and is a control device capable of monitoring the situation inside the furnace, which is difficult to observe visually or with a visible light camera, a boiler, and a method for acquiring a surveillance image of a boiler. And to provide a surveillance image acquisition program for boilers.

In order to solve the above problems, the control device, the boiler, the monitoring image acquisition method of the boiler, and the monitoring image acquisition program of the boiler adopt the following means.
The control device according to one embodiment of some embodiments of the present invention is a control device that monitors a boiler equipped with a furnace, and is an image that acquires a moving image of a monitored portion inside the furnace as data by an infrared camera. The acquisition unit and the captured moving image data acquired by the image acquisition unit are divided into a plurality of areas for each frame, image processing based on the brightness is performed for each of the areas, and the brightness of the processed image after the image processing is integrated for a predetermined time. It is provided with an image processing unit that acquires a transmission region transmitted to the monitoring target portion from the integrated image obtained in the above process, and a transmission image creation unit that creates a transmission image of the monitoring target portion based on a plurality of the transmission target portions. ..

According to this aspect, a moving image of the inside of a furnace taken by an infrared camera is acquired, this is divided into a plurality of areas for each frame, and image processing based on the brightness is performed for each area, and the brightness of the processed image is performed. From the integrated image obtained by accumulating the above for a predetermined time, the transmission area transmitted to the monitoring target part is acquired, and the transmission image of the monitoring target part is created from a plurality of transmission areas. It is possible to confirm the internal situation of the operating furnace with an image.
In the video taken by the infrared camera, the combustion state and fuel concentration in the furnace fluctuate irregularly, so the monitored part is blocked by flames, fuel, etc., and the monitored part is only momentarily or partially affected. I can't confirm. However, in this embodiment, since image processing is performed to acquire a transparent region from the integrated image and a transparent image is created, the transparent image from which the influence of short-time fluctuation components is removed improves visibility and makes the monitored portion clear. You can check.
This makes it possible to monitor the monitored part in the furnace even during operation. When monitoring the deposits on the monitored part, it is possible to determine whether or not the deposits are attached to the monitored part, and it is possible to prevent troubles caused by the deposits. be.

In the above aspect, the image processing unit may perform image processing so as to enhance the contrast for each of the regions.

According to this aspect, since the contrast is emphasized for each region, the difference in brightness can be further amplified to enhance the image of the region of the monitored portion that can be observed instantaneously. For example, if the overall contrast is emphasized, the contour of the monitored portion cannot be emphasized, but if the contrast is emphasized for each region as in this embodiment, the observed region of the monitored portion, that is, the contour of the transmission region is emphasized. Can be emphasized.

In the above aspect, the image processing unit acquires the transmission region based on the difference between the brightness of the integrated image and the value obtained by multiplying the average brightness of the processed image at the predetermined time by the number of frames at the predetermined time. May be good.

According to this aspect, since the transmission region is acquired based on the difference between the brightness of the integrated image and the value obtained by multiplying the average brightness of the processed image for a predetermined time by the number of frames, the influence of the fluctuation component due to the time fluctuation in the furnace is affected. Can be reduced.

In the above aspect, the image processing unit detects the presence or absence of vibration of the monitored portion, and when it detects that the monitored portion is vibrating, the image processing unit recognizes the vibration of the monitored portion and monitors the monitoring. The position of the target portion may be corrected.

According to this aspect, when it is detected that the monitored part is vibrating, the vibration of the monitored part is recognized and the position is corrected. Therefore, the integrated image when the monitored part is vibrating. In the above, there is a possibility that the monitored portion is exposed multiple times and the image is blurred, but this can be prevented and an integrated image without blurring can be obtained. As a result, the accuracy of image recognition can be improved even in the monitored portion where vibration is generated.

In the above aspect, the characteristics of the image of the teacher data created by marking the adhesion range of the deposit on the transparent image created by the transparent image creating unit are quantified, and the machine is based on the quantified teacher data. It may be provided with a learning unit that performs learning and creates a prediction model.

According to this aspect, the characteristics of the image of the teacher data created by marking the adhesion range of the deposit on the transparent image are quantified, and machine learning is performed based on the quantified teacher data to create a prediction model. Therefore, it is possible to quantitatively evaluate the state of the adhered deposits from the moving images of the monitored portion inside the furnace taken by the infrared camera, and predict the adhered range and the adhered amount.
In addition, since machine learning is performed using a transmission image that reduces the influence of fluctuation components due to time fluctuations in the furnace, the accuracy of machine learning is improved and the prediction accuracy of the prediction model is improved.

In the above aspect, even if the learning unit is provided, the teacher data created by classifying each region according to the image similarity is machine-learned for each region of the classified teacher data to create a prediction model. good.

According to this aspect, the teacher data created by classifying each area according to the image similarity is machine-learned for each area of the classified teacher data to create a prediction model, so that the learning efficiency is improved. , The prediction error can be reduced.

In the above aspect, the control device predicts the state of the deposit with reference to the prediction model for the transmission image of the captured moving image taken by the infrared camera, and notifies the prediction result. You may be prepared.

According to this aspect, the state of the deposit is predicted with reference to the prediction model for the transmitted image of the captured moving image taken by the infrared camera, and the prediction result is notified. Therefore, the monitored portion is monitored based on the captured moving image data. It is possible to quantitatively evaluate the state of deposits in.
In addition, by quantifying the state of adhesion of deposits to the monitored site, the operator can formulate a work plan based on the notified prediction results, so troubles caused by deposits can be prevented. It is possible.

In the above aspect, the prediction unit may monitor the prediction result of the state of the deposit and notify an alarm when the prediction result of the state of the deposit exceeds a predetermined threshold value.

According to this aspect, since the prediction result of the state of the deposit is monitored and an alarm is notified when the prediction result exceeds the threshold value, the state of the deposit progresses by the control device without constant monitoring by the operator. You can monitor what you are doing. This makes it possible to reduce the burden on the worker's monitoring work and prevent troubles caused by deposits.

In the above aspect, the learning unit may create the prediction model for each of a plurality of monitored parts of the furnace.

According to this aspect, since a prediction model is created for each of a plurality of monitored parts of the reactor, a different prediction model is created for each monitored part, and detailed prediction specific to each monitored part is performed. be able to.

The boiler according to one embodiment of some embodiments of the present invention comprises the control device according to any one of the above.

The method for acquiring a surveillance image of a boiler according to an aspect of some embodiments of the present invention includes an image acquisition step of acquiring a moving image of a monitored portion inside a furnace as data by an infrared camera, and an image acquisition step. The image processing step of dividing the captured moving image into a plurality of areas for each frame and performing image processing based on the brightness for each area, and the integration obtained by integrating the brightness of the processed image after the image processing for a predetermined time. The present invention includes a transmission region acquisition step of acquiring a transmission region transmitted from an image to the monitoring target portion, and a transmission image creation step of creating a transmission image of the monitoring target portion based on a plurality of the transmission target portions.

The boiler monitoring image acquisition program according to one aspect of some embodiments of the present invention includes an image acquisition step of acquiring a moving image of a monitored portion inside a furnace as data by an infrared camera, and an image acquisition step. The image processing step of dividing the captured moving image into a plurality of areas for each frame and performing image processing based on the brightness for each area, and the integration obtained by integrating the brightness of the processed image after the image processing for a predetermined time. The present invention includes a transmission region acquisition step of acquiring a transmission region transmitted from an image to the monitoring target portion, and a transmission image creation step of creating a transmission image of the monitoring target portion based on a plurality of the transmission target portions.

According to the present invention, since the transmitted image is created by performing image processing on the image taken by the infrared camera, it is possible to prevent the observation range from being obstructed by the floating matter in the furnace and to check the adhesion state of the deposit to the monitored portion. It can be confirmed clearly.
Further, according to the present invention, since machine learning is performed on the adhesion state of the deposit and a prediction model is created, the state of the deposit can be quantitatively evaluated based on the image taken by the infrared camera.

It is a schematic block diagram which shows one aspect of the boiler which concerns on some embodiments. It is a figure which showed the image by a visible light camera and the image by an infrared camera. It is a figure which showed the example which image-processed the whole image as a reference example. It is a schematic block diagram which shows one aspect of the transmission image of the monitored part and the image for teacher data which concerns on some embodiments. It is a flowchart which shows one aspect of the model construction method which concerns on some Embodiments. It is a graph which shows one aspect of the time transition of the ash adhesion amount of the monitored part of the furnace which concerns on some embodiments. It is a schematic block diagram which shows one aspect of the photographed image of the part to be monitored which concerns on some embodiments.

Hereinafter, an embodiment of a control device, a boiler, a boiler monitoring image acquisition method, and a boiler monitoring image acquisition program according to some embodiments of the present disclosure will be described with reference to the drawings.

FIG. 1 shows a schematic configuration of a control device, a boiler, a monitoring image acquisition method for a boiler, and a monitoring image acquisition program for a boiler according to the present embodiment.
As shown in FIG. 1, the pulverized coal-fired boiler (boiler) 10 is a conventional boiler and has a furnace 11 and a combustion device 12. The furnace 11 has a hollow shape of a square cylinder and is installed along the vertical direction, and the furnace wall constituting the furnace 11 is formed of a heat transfer tube.

The combustion device 12 is provided at the lower part of the furnace wall (heat transfer tube) constituting the furnace 11. The combustion device 12 has a plurality of combustion burners 21, 22, 23, 24, 25 mounted on the furnace wall. In this embodiment, the combustion burners 21, 22, 23, 24, 25 are arranged at equal intervals of four along the circumferential direction as one set, and five sets along the vertical direction, that is, It is arranged in 5 stages. The shape of the furnace 11, the number of combustion burners 21, 22, 23, 24, 25 in one stage, and the number of stages are not limited to this embodiment.

Each of the combustion burners 21, 22, 23, 24, and 25 constituting the combustion device 12 of the present embodiment has an oil nozzle capable of injecting oil fuel in the center and a fine powder fuel mixture outside the oil nozzle. It has a fuel nozzle that can be injected, a secondary air nozzle that can inject secondary air to the outside of the fuel nozzle, and a tertiary air nozzle that can inject tertiary air to the outside of the secondary air nozzle. There is. Therefore, at the time of starting the boiler, each combustion burner 21, 22, 23, 24, 25 injects oil fuel into the furnace 11 to form a flame, and then blows the fine fuel mixture and the secondary air and the tertiary air. It is injected into the furnace 11 to form a flame.

Further, as shown in FIG. 1, the flue 50 is connected to the upper part of the furnace 11, and the superheater 51 for recovering the heat of the exhaust gas as a convection heat transfer unit is connected to the flue 50. 52, reheaters 53, 54, and economizers (economizers) 55, 56, 57 are provided, and heat exchange is performed between the exhaust gas generated by combustion in the furnace 11 and water.

In the pulverized coal-fired boiler 10 configured as described above, the pulverized coal produced by the pulverized coal machine (not shown) is supplied to the combustion burners 21, 22, 23, 24, 25 together with the transport air. Further, the heated combustion air is supplied to each of the combustion burners 21, 22, 23, 24, and 25.

Then, the combustion burners 21, 22, 23, 24, 25 blow the pulverized fuel mixture in which the pulverized coal and the transport air are mixed into the furnace 11 and the combustion air into the combustion region of the furnace 11, at this time. By igniting, a flame swirling flow can be formed in this combustion region.

Then, the water supplied from the water supply pump (not shown) is preheated by the economizers 55, 56, 57, and then supplied to the steam drum (not shown) and supplied to each water pipe (not shown) of the furnace wall. It is heated to saturated steam and sent to a steam drum (not shown). Further, the saturated steam of the steam drum (not shown) is introduced into the superheaters 51 and 52 and superheated by the combustion gas. The superheated steam generated by the superheaters 51 and 52 is supplied to a power plant (for example, a turbine or the like) (not shown). Further, the steam taken out in the middle of the expansion process in the turbine is introduced into the reheaters 53 and 54, superheated again, and returned to the turbine. Although the furnace 11 has been described as a drum type (steam drum), the structure is not limited to this.

After that, the exhaust gas that has passed through the economizers 55, 56, 57 of the flue 50 has a denitration device (not shown) that removes harmful substances such as NOx by a catalyst, and an electrostatic precipitator removes particulate matter, and a desulfurization device. After the sulfur content is removed by the chimney, it is discharged into the atmosphere from the chimney.

An infrared camera 40 is installed in the furnace 11 so that the inside thereof can be photographed. The position of the infrared camera 40 in FIG. 1 is an example, and may be installed at any position as long as a part in the furnace 11 to be monitored (hereinafter referred to as a part to be monitored) can be photographed.
The infrared camera 40 captures a moving image of the inside of the furnace 11, and its frame rate is, for example, about 250 fps. Further, as the infrared camera 40, a medium-infrared camera having a wavelength of about 4 μm is preferably used.

A control device 30 is connected to the infrared camera 40. The infrared camera 40 is controlled by the control device 30.
The control device 30 is composed of, for example, a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), a computer-readable storage medium, and the like. As an example, a series of processes for realizing various functions are stored in a storage medium or the like in the form of a program, and the CPU reads this program into a RAM or the like to execute information processing / arithmetic processing. As a result, various functions are realized. The program is installed in a ROM or other storage medium in advance, is provided in a state of being stored in a computer-readable storage medium, or is distributed via a wired or wireless communication means. Etc. may be applied. The computer-readable storage medium is a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like.

The control device 30 includes an image acquisition unit 31, an image processing unit 32, a transparent image creation unit 33, a learning unit 34, and a prediction unit 35.
The image acquisition unit 31 acquires, as data, a moving image of a monitored portion inside the furnace 11 taken by the infrared camera 40.
The image processing unit 32 performs image processing, which will be described later, on the captured moving image data acquired by the image acquisition unit 31, and acquires a transparent region transmitted to the monitored portion.
The transparent image creating unit 33 creates a transparent image of the monitored portion based on the plurality of transparent regions acquired by the image processing unit 32.
The learning unit 34 performs machine learning based on the teacher data created from the transparent image of the monitored portion to create a prediction model. At this time, machine learning may create a plurality of prediction models for each of a plurality of transparent regions as needed.
The prediction unit 35 predicts the state of the deposit by referring to the prediction model linked to the target area with respect to the captured moving image data of the monitored portion photographed by the infrared camera 40, and notifies the outside.

In the pulverized coal-fired boiler 10, ash and the like are generated by combustion. The generated ash (clinker) adheres to the furnace wall of the furnace 11, the superheaters 51, 52, the reheaters 53, 54, the combustion burners 21, 22, 23, 24, 25, and the like. If ash is accumulated in the furnace 11 or adheres to a wide area, not only the operation of the furnace 11 is affected, but also the accumulated and solidified ash may fall, which may lead to damage to the furnace 11.
Therefore, in the present embodiment, the image taken by the camera is used, and image processing and learning are performed to monitor and predict the adhesion state of the deposit.

The monitoring of the furnace 11 using the control device 30 of the present embodiment will be described below.
First, an image processing method for reducing the influence of fluctuations in the furnace 11 will be described.
FIG. 2 shows an image taken by a visible light camera and an image taken by an infrared camera.
FIG. 3 shows an example in which the entire image is image-processed as a reference example.
The images of FIGS. 2 and 3 are images taken of the same monitored portion, FIG. 2 (a) is an image taken by a visible light camera, FIG. 2 (b) is an image taken by an infrared camera 40, and FIG. 2 (c) is an image taken. FIG. 2B shows an image in which the contrast of the image of FIG. 2B is emphasized, and FIG. 2D shows an image obtained by integrating a plurality of images of FIG. 2C.
As shown in FIG. 2A, even if the inside of the furnace 11 is photographed with a visible light camera, the radiation of the pulverized coal flame is strong and it is difficult to confirm the monitored portion.
As shown in FIG. 2B, when the inside of the furnace 11 is photographed by the infrared camera 40, the floating matter in the furnace 11 is photographed in white so as to block the monitored portion. Therefore, the monitored part can be observed only momentarily and partially.

Therefore, the image processing unit 32 performs the following image processing.
As shown by the dotted line in FIG. 2C, the image processing unit 32 performs a process of dividing the captured moving image data acquired by the image acquisition unit 31 into a plurality of small areas (small areas) for each frame.
Next, a process for enhancing the contrast (histogram smoothing process) is performed for each of the divided small areas. As a result, in a small region of the monitored portion that is not blocked by the suspended matter, the contrast is emphasized and a transmission region that has penetrated to the monitored portion can be obtained. In the transparent region, it is possible to confirm the contour of the monitored portion.
Here, since the position of the suspended matter is different for each frame, the position of the transmission region shown in FIG. 2C is also different. Therefore, the brightness of the image subjected to the histogram smoothing processing is integrated for a predetermined time for each small area of each frame, and the brightness of the integrated image (hereinafter referred to as the integrated image) and the small area of the frame corresponding to the predetermined time are integrated. For each, the difference from the value obtained by multiplying the average brightness of the image subjected to the histogram smoothing processing by the number of frames corresponding to a predetermined time is obtained. As a result, it is possible to connect only the transmission regions to reduce the influence of time fluctuations in the furnace 11 such as suspended matter, and obtain a transmission image (see FIG. 2D) in which only the monitored portion is clearly displayed. can.

On the other hand, when the image processing unit 32 performs histogram smoothing processing on the entire image without dividing into small areas, the contour of the monitored portion does not appear as shown as a reference example in FIG. 3, and a transparent area is obtained. I can't.

Next, a method of constructing a model for quantifying the amount of deposits adhered from the image obtained by image processing will be described.
FIG. 4 shows a transmission image of the monitored portion and an image for teacher data.
FIG. 4A is a transparent image of the monitored portion, and FIG. 4B is an image for teacher data of the monitored portion. Here, the teacher data is data corresponding to the result for the input data given for learning the pattern.
As shown in FIG. 2D, by performing image processing, a transparent image in which the monitored portion is clearly displayed can be obtained. Here, if the deposit is attached to the monitored portion, the deposit will appear white in the transmitted image as shown in FIG. 4 (a). However, there are some effects that cannot be removed even by image processing, such as the radiation of flames, and it is known that these effects also appear white like deposits.
Therefore, the operator determines whether or not the transparent image obtained by the image processing is an adhered substance. The operator confirms the transmitted image obtained by the image processing, and determines whether the white area is due to the deposit or the influence other than the deposit such as the radiation of the flame. The area determined to be an deposit by the operator is marked on the transmitted image as shown in FIG. 4 (b). The transparent image in which the deposits are marked in this way is used as the teacher data. At this time, the teacher data may be classified and created according to the image similarity of the area. In this case, the pattern is learned for each classified area. In this way, it is also effective to classify and create teacher data according to the image similarity of the area.

FIG. 5 shows a flowchart of the model construction method of the present embodiment.
First, we will explain the learning stage in which learning is performed when constructing a prediction model.
In step S501, the image acquisition unit 31 acquires (reads) the captured moving image of the monitored portion inside the furnace 11 captured by the infrared camera 40 as data.
Next, in step S502, the image processing unit 32 performs image processing on the captured moving image data acquired by the image acquisition unit 31, and acquires a transmission region transmitted to the monitored portion. Further, the transparent image creating unit 33 creates a transparent image of the monitored portion based on the plurality of transparent regions acquired by the image processing unit 32. The worker creates teacher data (learning data) from the transparent image of the monitored part.

Next, in step S503, the prediction model is updated.
The learning unit 34 performs machine learning based on the teacher data created from the transparent image of the monitored portion to create a prediction model. Here, for machine learning, a learning method such as Deep Learning is used. A typical method of deep learning includes CNN (Convolutional Neural Network) and the like.
The learning unit 34 quantifies the characteristics of the image from the brightness information of the data image of the teacher data, and learns the adhesion state of the deposit from the teacher data. As a result, the learning unit 34 creates a numerical model for determining the adhesion region of the deposit from the image information. This numerical model serves as a prediction model for the adhesion region of deposits.
In order to construct a highly accurate model, the processes from step S501 to step S503 are repeated in the learning stage, and the prediction model is updated.
Further, depending on the observation target, the learning may be divided according to the image similarity of the region. Since the monitored area has a form according to the original function and the form and characteristics of the deposits are different, the effect of improving the learning efficiency and the estimation accuracy by the numerical model by classifying in advance at the time of learning. Can be expected.
The prediction model thus created is saved (S504).
Further, inside the furnace 11, the state of adhesion of deposits differs depending on the position. Therefore, the prediction model is for each monitored part, such as a burner model created from images of combustion burners 21, 22, 23, 24, and 25, and an SH model created from images of superheaters 51 and 52. It is desirable to create and save it.

Next, we will explain the operation stage of monitoring using the constructed prediction model.
In step S511, the image acquisition unit 31 acquires (reads) the captured moving image of the monitored portion inside the furnace 11 captured by the infrared camera 40 as data. Then, the image processing unit 32 performs image processing on the captured moving image data acquired by the image acquisition unit 31, and acquires a transparent region transmitted to the monitored portion. Further, the transparent image creating unit 33 creates a transparent image of the monitored portion based on the plurality of transparent regions acquired by the image processing unit 32.
Next, in step S512, the prediction unit 35 refers to the prediction model saved in step S504 with respect to the transmitted image of the captured moving image data of the monitored portion captured by the infrared camera 40, and indicates the state of the deposit. Predict.
By referring to the prediction model corresponding to the monitored portion, the prediction unit 35 can automatically extract the adhesion region of the deposit from the transmitted image of the captured moving image data.
Next, in step S513, the prediction unit 35 notifies the outside of the prediction result of the adhesion region of the deposit. Examples of the method of notifying include a method of displaying on a display, turning on a lamp, and notifying by voice.

In this way, the amount of adhered matter can be quantified by using the area of the portion determined to be the adhered region of the adhered matter in the transmitted image of the captured moving image data. Since the amount of adhered matter is substantially proportional to the area of the adhered region, it is possible to estimate the amount of adhered matter from the area of the adhered region and perform a quantitative evaluation.

As the notification method, an alarm may be sounded in addition to the above.
FIG. 6 is a graph showing the time transition of the amount of ash adhered to the monitored portion of the furnace according to the present embodiment.
In FIG. 6, the vertical axis represents the amount of ash adhered and the horizontal axis represents time.
As shown in FIG. 6, the amount of ash adhered gradually accumulates over time. Further, a predetermined threshold value is set for the amount of ash adhered. The predetermined threshold value is a value set in advance by the operator based on various images and selecting an image of the adhesion amount which is the limit of the allowable range, and using the adhesion amount of the image as a predetermined threshold value. In the present embodiment, a predetermined threshold value is set to α.
When the amount of ash adhered, which increases with the passage of time, exceeds a predetermined threshold value α, the prediction unit 35 of the control device 30 sounds an alarm.
The operator can recognize that the ash is adhering by the notification of the alarm and take necessary countermeasures before the trouble occurs.

As described above, the control device, the boiler, the monitoring image acquisition method of the boiler, and the monitoring image acquisition program of the boiler according to the present embodiment have the following effects.
According to the present embodiment, a moving image of the inside of the furnace 11 taken by the infrared camera 40 is acquired, this is divided into a plurality of regions for each frame, and image processing based on the brightness is performed for each region for processing. From the integrated image obtained by integrating the brightness of the image for a predetermined time, the transmission area transmitted to the monitoring target part is acquired, and the transmission image of the monitoring target part is created from multiple transmission areas, so it can be confirmed visually or with a normal camera. It is possible to confirm the internal situation of the operating furnace 11 which cannot be performed by the image.
In the moving image taken by the infrared camera 40, since the combustion state and fuel concentration in the furnace 11 fluctuate unsteadily, the monitored portion is obstructed by flames, fuel, etc., and is monitored only momentarily or partially. The site cannot be confirmed. However, in the present embodiment, since image processing is performed to acquire a transparent region from the integrated image and a transparent image is created, visibility is improved by the transparent image from which the influence of short-time fluctuation components is removed, and the monitored portion is sharpened. Can be confirmed in.
This makes it possible to monitor the monitored portion in the furnace 11 even during operation. When monitoring the deposits on the monitored part, it is possible to determine whether or not the deposits are attached to the monitored part, and it is possible to prevent troubles caused by the deposits. be.

Further, according to the present embodiment, since the contrast is emphasized for each region, the difference in brightness can be further amplified to enhance the image of the region of the monitored portion that can be observed instantaneously. For example, if the overall contrast is emphasized, the outline of the monitored portion cannot be emphasized, but if the contrast is emphasized for each region as in the present embodiment, the observed region of the monitored portion, that is, the transmission region The outline can be emphasized.

Further, according to the present embodiment, since the transmission region is acquired based on the difference between the brightness of the integrated image and the value obtained by multiplying the average brightness of the processed image for a predetermined time by the number of frames, the fluctuation component due to the time fluctuation in the furnace 11 is obtained. The influence of can be reduced.

Further, according to the present embodiment, the characteristics of the image of the teacher data created by marking the adhesion range of the deposit on the transparent image are quantified, and machine learning is performed based on the quantified teacher data to obtain a prediction model. In order to create the data, it is possible to quantitatively evaluate the state of the adhered deposits from the moving images of the monitored portion inside the furnace 11 taken by the infrared camera 40, and predict the adhered range and the adhered amount.
In addition, the teacher data created by classifying each area according to the image similarity is machine-learned for each area of the classified teacher data to create a prediction model, which improves learning efficiency and reduces prediction errors. Can be reduced.
Further, since machine learning is performed using a transmission image in which the influence of fluctuation components due to time fluctuations in the furnace 11 is reduced, the accuracy of machine learning is improved and the prediction accuracy of the prediction model is improved.

Further, according to the present embodiment, the state of the deposit is predicted with reference to the prediction model for the transmitted image of the captured moving image taken by the infrared camera 40, and the prediction result is notified. Therefore, based on the captured moving image data. It is possible to quantitatively evaluate the state of deposits at the monitored site.
In addition, by quantifying the state of adhesion of deposits to the monitored site, the operator can formulate a work plan based on the notified prediction results, so troubles caused by deposits can be prevented. It is possible.

Further, according to the present embodiment, the prediction result of the state of the deposit is monitored, and an alarm is notified when the prediction result exceeds a predetermined threshold value. Therefore, the control device 30 is attached without constant monitoring by the operator. It is possible to monitor the progress of the kimono condition. This makes it possible to reduce the burden on the worker's monitoring work and prevent troubles caused by deposits.

Further, according to the present embodiment, since the prediction model is created for each of the plurality of monitored parts of the reactor 11, different prediction models are created for each monitored part, and the detailed details peculiar to each monitored part are created. You can make predictions.

In the above-described embodiment, the vibration in the furnace is not examined, but in the present embodiment, appropriate image processing is performed even in the vibrating monitored portion. Since other points are the same as those of the above-described embodiment, the same reference numerals are given to the same configurations, and the description thereof will be omitted.

FIG. 7 shows a photographed image of the monitored portion according to the present embodiment.
Each image of FIG. 7 is an image obtained by taking the same monitored portion, FIG. 7 (a) is an image taken by the infrared camera 40, FIG. 7 (b) is an integrated image of the monitored portion, and FIG. 7 (c). Shows an integrated image obtained by performing blur correction processing on a monitored portion.
The inside of the furnace 11 may be affected by the flow of gas or the like during operation, and for example, the superheaters 51 and 52 and the reheaters 53 and 54 may vibrate.
Even if the monitored portion vibrates, as shown in FIG. 7A, there is no particular problem in the image taken by the infrared camera 40.
However, as shown in FIG. 7 (b), in the integrated image obtained by integrating a plurality of images of FIG. 7 (a), the monitored portion may be exposed in multiple layers, which may cause a problem that the image is blurred.

Therefore, in the present embodiment, before the image processing unit 32 integrates the images, the contour of the monitored portion is automatically recognized, the vibration is recognized, and the position is corrected.
When the image processing unit 32 corrects the position (correction of blurring), it is possible to obtain an integrated image in which the outline of the monitored portion is clearly displayed as shown in FIG. 7 (c). However, by correcting the blur, the background of the monitored part will be blurred.

As described above, the control device, the boiler, the monitoring image acquisition method of the boiler, and the monitoring image acquisition program of the boiler according to the present embodiment have the following effects.
According to the present embodiment, when it is detected that the monitored part is vibrating, the vibration of the monitored part is recognized and the position is corrected. Therefore, the integration when the monitored part is vibrating is performed. There is a risk that the monitored portion will be exposed multiple times in the image and the image will be blurred, but this can be prevented and an integrated image without blurring can be obtained. As a result, the accuracy of image recognition can be improved even in the monitored portion where vibration is generated.

Although each embodiment of the present invention has been described in detail with reference to the drawings, the specific configuration is not limited to this embodiment.
For example, in each of the above-described embodiments, the adhesion region (adhesion amount) of the deposit is predicted, but the operation timing of the suit blower may be determined based on the predicted adhesion amount of the deposit.
By determining the operation timing of the suit blower using the prediction model, it is possible to improve the operation of the suit blower, reduce the amount of steam used in the suit blower, and increase the amount of power generation. Become. In addition, erosion caused by excessive operation of the suit blower can be suppressed.

10 Fine charcoal-fired boiler (boiler)
11 Fire furnace 12 Combustion device 21, 22, 23, 24, 25 Combustion burner 30 Control device 31 Image acquisition unit 32 Image processing unit 33 Transmission image creation unit 34 Learning unit 35 Prediction unit 40 Infrared camera 50 Flue 51, 52 Superheater ( Super heater)
53, 54 Reheater 55, 56, 57 Economizer

Claims (12)

A control device that monitors boilers equipped with a furnace.
An image acquisition unit that acquires the video of the monitored part inside the furnace with an infrared camera as data,
The captured moving image data acquired by the image acquisition unit is divided into a plurality of areas for each frame, image processing based on the brightness is performed for each area, and the brightness of the processed image after the image processing is integrated for a predetermined time. An image processing unit that acquires a transparent region transmitted from the integrated image to the monitored portion, and
A transparent image creation unit that creates a transparent image of the monitored portion based on the plurality of transparent regions,
A control device equipped with. The control device according to claim 1, wherein the image processing unit performs image processing so as to enhance the contrast in each of the regions. Claim 1 or claim 1, wherein the image processing unit acquires the transmission region based on the difference between the brightness of the integrated image and the value obtained by multiplying the average brightness of the processed image at the predetermined time by the number of frames at the predetermined time. Item 2. The control device according to item 2. The image processing unit detects the presence or absence of vibration of the monitored portion, and when it detects that the monitored portion is vibrating, the image processing unit recognizes the vibration of the monitored portion and positions the monitored portion. The control device according to any one of claims 1 to 3, wherein the amendment is made. The characteristics of the image of the teacher data created by marking the adhesion range of deposits on the transparent image created by the transparent image creating unit are quantified, and machine learning is performed based on the quantified teacher data. The control device according to any one of claims 1 to 4, further comprising a learning unit for creating a prediction model. Claims 1 to claim 1 include a learning unit for creating a prediction model by performing machine learning for each of the classified areas of the teacher data created by classifying each of the areas according to the degree of image similarity. The control device according to any one of 4. The fifth or sixth aspect of the present invention, wherein the transmission image of the captured moving image taken by the infrared camera is provided with a prediction unit for predicting the state of deposits with reference to the prediction model and notifying the prediction result. Control device. The control device according to claim 7, wherein the prediction unit monitors the prediction result of the state of the deposit and notifies an alarm when the prediction result of the state of the deposit exceeds a predetermined threshold value. The control device according to any one of claims 5 to 8, wherein the learning unit creates the prediction model for each of the plurality of monitored parts of the furnace. A boiler provided with the control device according to any one of claims 1 to 9. The image acquisition process to acquire the video of the monitored part inside the furnace with an infrared camera as data,
An image processing step of dividing the captured moving image acquired in the image acquisition step into a plurality of regions for each frame and performing image processing based on the brightness for each of the regions.
A transmission region acquisition step of acquiring a transmission region transmitted to the monitored portion from the integrated image obtained by integrating the brightness of the processed image after image processing for a predetermined time, and
A transmission image creation step of creating a transmission image of the monitored portion based on the plurality of transmission regions, and a transmission image creation step.
Boiler surveillance image acquisition method. An image acquisition step to acquire the video of the monitored part inside the furnace with an infrared camera as data,
An image processing step in which the captured moving image acquired in the image acquisition step is divided into a plurality of areas for each frame and image processing is performed based on the brightness in each of the areas.
A transmission area acquisition step for acquiring a transmission area transmitted to the monitored portion from the integrated image obtained by integrating the brightness of the processed image after image processing for a predetermined time, and
A transparent image creation step for creating a transparent image of the monitored portion based on the plurality of transparent regions, and a transparent image creation step.
Boiler surveillance image acquisition program.

Images