JP6459550B2 - Anomaly detection method and anomaly detection apparatus for mechanical equipment - Google Patents

Description

  The present invention relates to a method and apparatus for detecting an abnormality of a mechanical facility using an image.

Some of the machinery and equipment that make up the steelmaking process are equipped with many identical devices, and some of the same devices may become unusable due to damage during use. If it is not replaced, proper production cannot be continued.
However, since the mechanical equipment constituting the iron making process is equipped with many identical devices, it is very expensive to replace all of them at the same time.
Therefore, it is necessary to detect and replace only the equipment that has become unusable at an early stage. However, the equipment maintenance personnel are heavily burdened with the discovery of the unusable equipment.
As described above, it is an urgent issue to reduce the labor for the equipment maintenance personnel to find the equipment that has become unusable.
As for abnormal inspection of products, when detecting abnormal products among products that are produced in large quantities, images of products etc. may be collected by a camera, and the images may be processed by a computer. is there.

  For example, when detecting an abnormality in a product, a plurality of images of normal products are collected, an average image (hereinafter referred to as “average image”) is constructed from the collected images, and the average image and product There is a method in which a normal range or an abnormal range is determined by comparing with the image of the product, and a product that has entered the normal range or a product that has not entered the abnormal range is determined to be normal.

  In Patent Document 1, in a non-defective image group acquired from a plurality of non-defective products, alignment is performed, and an average luminance and a standard deviation are calculated for each pixel having the same coordinates, thereby inspecting the product. In addition, an image processing method is disclosed in which registration is performed on an image acquired from the above, and for each pixel having the same coordinates, the quality is determined by an evaluation formula defined using an average and a standard deviation of the luminance.

  However, using the method disclosed in Patent Document 1, it is difficult to perform abnormality determination of the same equipment that constitutes a mechanical facility used in, for example, a steel manufacturing process.

  First, since a normal product is almost uniform immediately after production to be inspected, it is difficult to make an error even if it is determined as an abnormal product if there is a unique part in the product image. However, the equipment that makes up the machinery and equipment that produces the product may change due to aging, wear, etc., and is not necessarily uniform, so even if there is a unique part in the image of the equipment, it is not necessarily determined to be abnormal. This is because there are many cases that must not.

  Second, the environment around the equipment that constitutes the machine equipment that produces the product often changes greatly, and even if the equipment is normal, the change in the environment around the equipment that constitutes the machine equipment that produces the product is the equipment. May cause a unique part of the image. Therefore, even if a unique portion exists in the image of the device, it is often not necessarily determined as abnormal.

  Patent Document 2 stores a plurality of image data determined to be non-defective among images obtained by imaging an inspection object, and performs inspection based on the stored plurality of image data determined to be non-defective There is disclosed a method for calculating a threshold value for performing pass / fail determination of an object and determining whether or not the imaged inspection object is a non-defective product.

However, even if the method disclosed in Patent Document 2 is used, it is difficult to carry out abnormality determination of the same equipment that constitutes the mechanical equipment used in the above-described steel manufacturing process, for example.
Because, unlike the case of inspecting products with basically uniform quality, the environment around the equipment that makes up the machinery and equipment that produces the products often changes greatly, and the equipment that makes up the machinery and equipment that produces the products. If the threshold value is not changed according to the surrounding environment, so-called undetected cases in which an abnormal image is determined as a normal image occur frequently.

  Patent Document 3 discloses a defect threshold for receiving and storing a plurality of images constituting a non-defective image group, and detecting a defective portion of an inspection object based on the plurality of stored images. In addition, a determination threshold value for performing pass / fail determination is set, a defective product image determined to be a defective product is stored, and when an input of an image acquired by newly imaging an inspection object is received, an input is received. The non-defective product learning process is executed using a plurality of stored images including the set image, at least the defect threshold is reset, the defect portion is detected again based on the reset defect threshold, and the set determination threshold is set A method is disclosed in which it is determined whether or not a defective product image stored based on the above is a defective product, and the result of determination of good or bad is displayed.

However, even if the method disclosed in Patent Document 3 is used, it is difficult to carry out abnormality determination of equipment that constitutes the mechanical equipment used in the steel manufacturing process described above, for example.
This is because, unlike the case of inspecting products having basically uniform quality, the number of defective product images determined to be defective according to the environment around the equipment constituting the machine equipment that produces the product is enormous. It is difficult to memorize all this. Even if a defective product image can be stored, it becomes difficult to distinguish it from a normal image similar to the defective product image, and so-called overdetection that determines a normal image as an abnormal image occurs frequently.

JP 2005-265661 A JP 2013-224833 A JP 2013-142558 A

  The present invention is not limited to the mechanical equipment used in the steel manufacturing process, and even if there is a change such as aging or wear of equipment constituting the mechanical equipment or a change in the environment surrounding the mechanical equipment, the mechanical equipment It is an object of the present invention to provide a detection method and apparatus with low so-called “undetected” in which an abnormality of the equipment constituting the device is overlooked and so-called “overdetection” that determines an abnormality when there is no abnormality.

  As a result of earnest research and development, the inventors have made an apparatus that is an abnormality detection target (hereinafter referred to as an “abnormality detection target apparatus”) when an image of an apparatus that constitutes mechanical equipment is collected by an imaging device such as a camera. It has been found that the incidence of undetected and overdetection can be significantly reduced by spraying a fluid such as water and irradiating light from a plurality of different directions simultaneously with imaging.

Furthermore, the inventors have photographed mechanical equipment from different directions using a plurality of imaging devices, and even when a three-dimensional image of the mechanical equipment is collected, a fluid such as water is sprayed on the abnormality detection target device, It has been found that it is effective to irradiate the abnormality detection target device with light from different directions simultaneously with imaging, and it is also effective when the imaging device is a color camera.
The present invention has been made based on these findings, and the gist thereof is as follows.

(1) A first aspect of the present invention is a machine facility abnormality detection method for detecting an abnormality of a device provided in a steel manufacturing machine facility, and includes an image of an abnormality detection target device (hereinafter, “device image”). )
Spraying fluid on the anomaly detection target device;
Irradiating the abnormality detection target device with light from a plurality of different directions;
Capturing an image of the abnormality detection target device simultaneously with irradiating the abnormality detection target device with light from a plurality of different directions; and
A group of steps taken through
From a plurality of the device images (hereinafter referred to as “image group”),
Searching for a device image (hereinafter referred to as “abnormal image”) having an abnormal image portion (hereinafter referred to as “abnormal portion”);
Deleting abnormal images from the image group and configuring the deleted image group;
A group of steps constituting the pixel information of the representative image having the average value x _i and the standard deviation σ _i of the corresponding pixel i in each image of the post-deletion image group;
A device image of the abnormality detection target device at a certain time (hereinafter referred to as a “device image at a certain time”),
Spraying fluid on the anomaly detection target device;
Irradiating light from a plurality of different directions to the abnormality detection target device and simultaneously collecting the device image;
A group of steps obtained through
The abnormality determination of the device image at the certain time
When the luminance y _i of the pixel i of the device image is k as a real number,
x _i −kσ _i ≦ y _i ≦ x _i + kσ _i (in the case where (Equation 1) is not satisfied)
Counting the number n of pixels determined to be abnormal pixels and adjacent to each other;
Determining that the image is an abnormal image when n is a predetermined number N or more;
The abnormality detection target device is a sintering machine .

(2) In the above aspect (1), the real number k may be a real number of 1 or more.

(3) In the above aspect (1) or (2), when the determination step is performed a plurality of times, the total number of determinations by the determination step (hereinafter referred to as “total determination number”) M and the determination step Counting the total number of determinations determined as abnormal images (hereinafter referred to as “abnormal determination number”) m;
Calculating P = m / M (Equation 2) from the total determination number M and the abnormality determination number m for the abnormality detection target device;
If P is greater than or equal to a predetermined value, the image processing apparatus may further include a step group including a step of final determination as an abnormal image.

(4) In the aspect of the above (3), when there are a plurality of imaging units that capture images of the abnormality detection target device simultaneously with the step of irradiating the abnormality detection target device with light from a plurality of different directions,
A step group constituting an image group for each imaging unit;
In each image group, a step group that configures a post-deletion image group in which an abnormal image is deleted and configures pixel information of each representative image corresponding to each post-deletion image group;
May further be included.

(5) In the aspect described in any one of (1) to (4) above, a direction in which an image of the abnormality detection target device is captured is a normal line of a plane including the inspection target surface of the abnormality detection target device. The angle between 0 ° to 60 ° and the angle between the direction of irradiating the light and the normal of the plane including the inspection target surface of the abnormality detection target device may be set to 30 ° to 60 °.

(6) In the aspect described in any one of (1) to (5) above, the fluid may be water.

(7) In the aspect described in any one of (1) to (5) above, the fluid may be steam.

(8) In the above aspect (6) or (7), after the fluid is sprayed, a gas may be sprayed on the abnormality detection target device.

(9) In the aspect according to any one of (6) to (8), in the step of capturing an image of the abnormality detection target device, the imaging device has detection sensitivity in a wavelength band of 1.1 μm or less. An imaging device may be used, and an optical filter that cuts light with a wavelength of 0.7 μm or less and an optical filter that cuts light with a wavelength of 0.95 to 1.0 μm may be attached to the imaging device in an overlapping manner.

( 10 ) In addition, another aspect of the present invention is an abnormality detection apparatus for machine equipment for detecting an abnormality of the equipment provided in the steel manufacturing machine equipment, and for each abnormality detection target device,
A fluid spraying part for spraying fluid on the abnormality detection target device;
An illumination unit that emits light from a plurality of different directions to the abnormality detection target device;
An imaging unit that collects an image of the abnormality detection target device (hereinafter, referred to as “device image”) at the same time that the illumination unit irradiates light;
A representative image pixel information constituting unit constituting representative image pixel information for a plurality of device images (hereinafter referred to as “image group”);
A determination unit that determines whether or not there is an abnormality in a device image at a certain time of the abnormality detection target device (hereinafter referred to as a “device image at a certain time”);
The representative image each pixel information constituting unit,
A post-deletion image group constituting unit that deletes a device image (hereinafter, referred to as “abnormal image”) having an abnormal image part (hereinafter referred to as “abnormal part”) from the image group and constitutes a post-deletion image group; ,
An average value x _i of the luminance of the corresponding pixel i of each image in the post-deletion image group and a standard image luminance component that is used as each pixel information of the standard deviation σ _i representative image;
Have
The determination unit
When the luminance y _i of the pixel i of the device image at a certain time is k as a real number,
x _i −kσ _i ≦ y _i ≦ x _i + kσ _i ... (Expression 1)
Means for counting the number n of pixels determined to be abnormal pixels and adjacent to each other;
Means for determining that the image is an abnormal image when n is a predetermined number N or more;
Have a, the abnormality detecting target device is a sintering machine.

( 11 ) In the above aspect ( 10 ), the real number k may be a real number of 1 or more.

( 12 ) In the above aspect ( 10 ) or ( 11 ), the determination unit includes a plurality of determination units,
A counting unit that counts the total number of determinations (hereinafter referred to as “total number of determinations”) M and the total number of determinations determined as abnormal images (hereinafter referred to as “abnormality determination numbers”) m in the determination unit. When,
A calculation unit for calculating P = m / M (Expression 2) from the total determination number M and the abnormality determination number m;
If the P is equal to or greater than a predetermined value, a final determination unit that finally determines an abnormal image;
You may further have the comprehensive determination part which has.

( 13 ) In the aspect of ( 12 ), when the abnormality detection target device has a plurality of imaging units that irradiate light from a plurality of different directions and simultaneously capture images of the abnormality detection target device,
An image collection unit that constitutes an image group for each imaging unit;
Each image group may further include a representative image pixel information configuration unit that configures the post-deletion image group from which the abnormal image is deleted and configures the representative image pixel information corresponding to each post-deletion image group. .

( 14 ) In the aspect described in any one of ( 10 ) to ( 13 ) above, a direction in which an image of the abnormality detection target device is captured is a normal line of a plane including the inspection target surface of the abnormality detection target device. The angle is 0 ° to 60 °, and the angle between the direction of irradiating the light and the normal line of the plane including the inspection target surface of the abnormality detection target device may be 30 ° to 60 °.

( 15 ) In the aspect according to any one of ( 10 ) to ( 14 ), the fluid may be water.

( 16 ) In the aspect described in any one of ( 10 ) to ( 14 ), the fluid may be air.

( 17 ) In the above aspect ( 15 ) or ( 16 ), after the fluid is sprayed, a gas may be sprayed on the abnormality detection target device.

( 18 ) In the aspect described in any one of ( 15 ) to ( 17 ), in the step of capturing an image of the abnormality detection target device, the imaging device has detection sensitivity in a wavelength band of 1.1 μm or less. An imaging device may be used, and an optical filter that cuts light with a wavelength of 0.7 μm or less and an optical filter that cuts light with a wavelength of 0.95 to 1.0 μm may be attached to the imaging device in an overlapping manner.

  By using the present invention, even if there is a change such as aging or wear of the equipment constituting the mechanical equipment, or a change in the environment surrounding the mechanical equipment, the abnormality of the equipment constituting the mechanical equipment is not detected or overdetected. There is a remarkable effect that a very appropriate judgment can be made with few.

It is a flowchart explaining the procedure of the method of this invention. It is a figure explaining the system configuration | structure of the abnormality detection apparatus of the mechanical installation which implements the method of this invention. It is a figure explaining the abnormality detection apparatus (imaging apparatus vertical type) of the mechanical installation which implements the method of this invention. FIG. 3A is a diagram viewed from the upper side of the abnormality detection target device, and FIG. 3B is a diagram viewed from the AA cross section of FIG. It is a figure explaining the abnormality detection apparatus (imaging device horizontal type) of the mechanical installation which implements the method of this invention. FIG. 4B is a diagram viewed from the upper side of the abnormality detection target device, and FIG. 4A is a diagram viewed from the BB cross section of FIG. It is a figure explaining the structure method of each pixel information of the representative image of this invention. It is a figure explaining the form which implements the method or apparatus of the present invention. FIG. 6A is a diagram when there is one imaging unit, and FIG. 6B is a diagram when there are two imaging units.

(First embodiment)
1st Embodiment of this invention is an abnormality detection apparatus of a mechanical installation.
FIG. 2 is a diagram for explaining the system configuration of an apparatus for detecting an abnormality of mechanical equipment that implements the method of the present invention. FIG. 2A shows a system configuration in a case where pixel information of each representative image is configured, and FIG. 2B shows a system configuration in a case where a device image at a certain time is determined.
In FIG. 2, reference numeral 7 denotes an image collecting unit, which includes a fluid spraying unit 3, an illumination unit 4, and an imaging unit 2.

First, a description will be given with reference to FIG. 2A, the output of the imaging unit 2 is connected to the terminal A, and the device image captured by the imaging unit 2 is sent to the representative image pixel information configuration unit 5 via the terminal A. The constructed representative image pixel information is sent to the determination unit 6.
Next, a description will be given with reference to FIG. In FIG. 2B, the output of the imaging unit is connected to the terminal B, and the device image captured by the imaging unit 2 is sent to the determination unit 6 via the terminal B, and abnormality determination is performed.

Next, with reference to FIG. 3 and FIG. 4, a description will be given of an abnormality detection method for machine equipment that implements the method of the present invention.
FIG. 3 is a diagram for explaining an apparatus for detecting an abnormality of mechanical equipment (an image pickup apparatus vertical type) that implements the method of the present invention.
FIG. 4 is a diagram for explaining an apparatus for detecting an abnormality of mechanical equipment (imaging apparatus horizontal type) that implements the method of the present invention.
3 differs in that the imaging device is installed in the vertical direction, and FIG. 4 differs in that the imaging device is installed in the horizontal direction.

In FIG. 3 or FIG. 4, 1 is an abnormality detection target device, 10 is a plane normal including the inspection target surface of the abnormality detection target device, 20 is an imaging direction, 21 is a first imaging unit, 22 is a second imaging unit, 23 is the third imaging unit, α ₁ is the first imaging unit installation angle, α ₂ is the second imaging unit installation angle, α ₃ is the third imaging unit installation angle, 30 is the fluid spraying direction, and 31 is the first fluid blowing unit. hook, the second fluid吹掛unit 32, 40 is the light emission direction, the first illumination unit 41, the second illumination section 42, the third illumination unit 43, the fourth illumination part 44, theta ₁ is The first illumination unit installation angle, θ ₂ is the second illumination unit installation angle, θ ₃ is the third illumination unit installation angle, and θ ₄ is the fourth illumination unit installation angle.

[Fluid spray part]
The fluid spraying unit 3 is a device that sprays fluid onto the abnormality detection target device 1, and the fluid is sprayed in the direction of the fluid spraying direction 30.
The reason why the fluid is sprayed on the abnormality detection target device is to blow off dust and dirt attached to the device. The environment surrounding steel production machinery is extremely poor, and equipment is often contaminated with oil.
Therefore, it is possible to remove the dirt by spraying the fluid and prevent the dirt from appearing as a unique portion in the device image. The pressure of the fluid sprayed is a pressure that does not damage the device.
In FIG. 3 or FIG. 4, it corresponds to the fluid spray direction 30, the first fluid spray portion 31, and the second fluid spray portion 32, and the plane normal 10 including the inspection target surface of the abnormality detection target device 1. Equipped in parallel directions.
The timing of spraying the fluid is preferably immediately before the device image is collected by the imaging unit (the first imaging unit, the second imaging unit, and the third imaging unit), and water, air, and the like can be used as the fluid. Desirably, the fluid is clean.

[Lighting part]
The illumination unit 4 is a device that irradiates the abnormality detection target device 1 with light from a plurality of different directions.
The reason for irradiating light is to make the brightness of the device uniform.
The environment surrounding steel production machinery is poor and often has some shade. This is because it is possible to prevent the shade generated in the device by irradiating light from appearing as an abnormal portion of the device image.

As for the direction which irradiates light to the abnormality detection object apparatus 1, it is desirable that it is at least two directions. This is because shadows may not be sufficiently removed in the case of irradiation from one direction.
In FIG. 3 or FIG. 4, it corresponds to the irradiation direction 40 of light, the 1st illumination part 41, the 2nd illumination part 42, the 3rd illumination part 43, and the 4th illumination part 44, and the inspection object surface of the abnormality detection target apparatus 1 is shown. The first illumination unit installation angle θ ₁ , the second illumination unit installation angle θ ₂ , the third illumination unit installation angle θ ₃ , and the fourth illumination unit installation angle θ ₄ are respectively set with respect to the normal line 10 including the plane. It is installed.
The first illumination unit installation angle θ ₁ , the second illumination unit installation angle θ ₂ , the third illumination unit installation angle θ ₃ , and the fourth illumination unit installation angle θ ₄ are planes that include the inspection target surface of the abnormality detection target device. It is desirable that the angle is 30 ° to 60 ° with respect to the normal 10. This is because taking such an angle suppresses the occurrence of shadows and makes the device image uniform.

[Imaging section]
The imaging unit 2 is a device that collects device images at the same time as irradiating light.
The reason why the device image is collected simultaneously with the light irradiation is that the generation of shadows is suppressed and the device image becomes uniform.
A plurality of device images captured by the imaging unit 2 form an image group.

In FIG. 3 or FIG. 4, corresponding to the imaging direction 20, the first imaging unit 21, the second imaging unit 22, and the third imaging unit 23, with respect to the plane normal 10 including the surface to be inspected of the abnormality detection target device 1. The first imaging unit installation angle α ₁ , the second imaging unit installation angle α ₂ , and the third imaging unit installation angle α ₃ are respectively installed.
Further, the first imaging unit installation angle α ₁ is not shown in FIG. 3 or FIG. 4, but α ₁ = 0 °. Thus, the first imaging unit installation angle is often parallel to the normal line 10 of the plane including the inspection target surface of the abnormality detection target device 1. The first imaging unit installation angle α ₁ , the second imaging unit installation angle α ₂ , and the third imaging unit installation angle α ₃ are preferably 0 ° to 60 °.
A normal camera can be used as the imaging unit.
Furthermore, in order to cut light caused by water, an optical filter that cuts light having a wavelength of 0.7 μm or less and an optical filter that cuts light having a wavelength of 0.95 to 1.0 μm are attached to the imaging unit. , More clear images can be taken.

[Representative image pixel information component]
This is a device that configures pixel information of each representative image based on an image group that is a plurality of device images captured by the imaging device.
The representative image pixel information about the collected image group of the abnormality detection target device is configured.
The representative image pixel information configuration unit will be described with reference to FIG.
FIG. 5A is a diagram depicting an image of the post-deletion image group, and FIG. 5B is a diagram depicting an image of each pixel information of the representative image, and FIG. 5A and FIG. ) Represents a pixel.
The post-deletion image group in FIG. 5A is obtained by deleting a device image having an abnormal part from the image group, and has a plurality of device images.
Therefore, for example, the average value x _i and the standard deviation σ _i are calculated for the luminance of the i-th pixel of each device image.
The information of the i-th pixel in the pixel information of each representative image in FIG. 5B is the previously calculated average value x _i and standard deviation σ _i .
Such an operation is performed for all the pixels, and each pixel information of the representative image is configured.

[Determining part]
The determination unit 6 is a device that determines abnormality of the device image at a certain time.
If the difference between the device image and the representative image of the device is greater than or equal to a predetermined value, the device is determined to be abnormal.
For example, when the luminance y _i of the pixel i of the device image at a certain time is k as a real number,
x _i −kσ _i ≦ y _i ≦ x _i + kσ _i (Equation 1)
Is not satisfied, it is determined as an abnormal pixel.
For example, in the case of k = 3, when the standard deviation σ _i of the luminance of the pixel deviates from the range of −3σ _i to + 3σ _i from the average value x _i, it is determined as an abnormal pixel, and the adjacent abnormal pixel The number n is counted, and when the number is a predetermined number N or more, the image is determined to be an abnormal image.
Here, the reason why the adjacent abnormal pixels are counted is that abnormal pixels that are not adjacent are noise, and so if it is determined as an abnormal image based on this, so-called overdetection increases.
The value of k can be determined on the basis of actual results, and it is preferable that the value of k be determined by correcting each time the present invention is implemented for each abnormality detection target device.

[Comprehensive judgment section]
The overall determination unit 8 determines the total number of determinations (hereinafter referred to as “total determination number”) M obtained from each determination unit and the determination number determined as an abnormal image (hereinafter referred to as “abnormal determination number”) m. It is a device that counts and makes a final determination from the total determination number M and the abnormality determination number m.
The overall determination unit 8 is an apparatus used when, for example, a plurality of determination units 6 are provided.
The case of having a plurality of determination units 6 includes 1) the case where the image collection unit 7 and the representative image pixel information configuration unit 5 have the same determination unit 6 but have different determination criteria, and 2) the image. A case where a plurality of collection units 7, representative image pixel information configuration units 5 and determination units 6 are provided is included.
By having the comprehensive determination unit as described below, there is a remarkable effect that it is possible to perform abnormality determination with less overdetection.

The comprehensive determination unit includes a counting unit, a calculation unit, and a final determination unit.
In FIG. 2, reference numeral 8 denotes an overall determination unit, which includes a counting unit 81, a calculation unit 82, and a final determination unit 83.
[Counter]
Count the total number of determinations M and the number of abnormality determinations m.
Basically, the total number of determinations M is the number of determination units 6. However, when some of the determination units 6 are not used, the total number of determination units 6 is obtained by subtracting the number of determination units 6 that are not used from the total number of determination units 6. .
[Calculation section]
From the total determination number M and the abnormality determination number m, P = m / M (Expression 2)
Calculate
[Final judgment part]
If P is greater than or equal to a predetermined value, it is finally determined as an “abnormal image”.
A value of 0.3 to 0.8 is used as the value given in advance. Desirably, a value of 0.5 to 0.7 is used.

<Configuration example>
For example, a case where “0.6” is used as a predetermined value will be described.
[When there are two judgment units]
Since there are two determination units, when M = 2 and m = 2, P is 0.6 or more. That is, only when two are determined to be abnormal images, the final determination is “abnormal image”, and when only one of them is determined to be “abnormal image”, the final determination is not “abnormal image”.
[When there are three judgment units]
Since there are two determination units, when M = 3 and m = 2, P is 0.6 or more.
That is, when any two are determined to be abnormal images, the final determination is “abnormal image”.

(Second Embodiment)
The second embodiment of the present invention is a method for detecting an abnormality of a device using the apparatus described in the first embodiment, and includes the following steps.

[Step Group 1] A step group for collecting device images.
A plurality of collected device images form an image group.
[Step 1-1] This is a step of spraying fluid on the abnormality detection target device.
Spray the fluid on the device to be detected. The reason why the fluid is sprayed on the abnormality detection target device is to blow off dust and dirt attached to the device. Although the environment surrounding the steel manufacturing machinery and equipment is remarkably changed and the equipment is often soiled with oil or the like, it is possible to remove the dirt by spraying a fluid and prevent the dirt from appearing as a unique part in the equipment image. The pressure of the fluid to be sprayed is a pressure higher than the atmospheric pressure, and is a pressure that does not damage the device.

  When the apparatus described in FIG. 3 or FIG. 4 is used, the first fluid spraying part 31 and the second fluid spraying part 32 execute Step 1-1. The first fluid spraying part 31 and the second fluid spraying part 32 are equipped in a direction parallel to the normal line 10 of the plane including the inspection target surface of the abnormality detection target device 1. This is because the fluid is preferably sprayed immediately before the device image is captured. The fluid used is water, air or the like.

[Step 1-2] This is a step of irradiating the abnormality detection target device with light from a plurality of methods.
The reason for irradiating light is to make the brightness of the device uniform. This is because the environment surrounding the steel manufacturing machine facility changes remarkably, but by irradiating light, it is possible to prevent the change in brightness around the device from appearing as a specific part of the device image.
It is desirable that the light irradiation angle is an angle that forms an angle of 30 ° to 60 ° with respect to the normal 10 of the plane including the inspection target surface of the abnormality detection target device. This is to suppress the occurrence of shadows and make the brightness of the device uniform.

When the apparatus described in FIG. 3 or FIG. 4 is used, the light irradiation direction 40, the first illumination unit 41, the second illumination unit 42, the third illumination unit 43, and the fourth illumination unit 44 are step 1- 2, the first illumination unit 41, the second illumination unit 42, the third illumination unit 43, and the fourth illumination unit 44, with respect to the normal 10 of the plane including the inspection target surface of the abnormality detection target device 1, The first illumination unit installation angle θ ₁ , the second illumination unit installation angle θ ₂ , the third illumination unit installation angle θ ₃ , and the fourth illumination unit installation angle θ ₄ are respectively installed.
The first illumination unit installation angle θ ₁ , the second illumination unit installation angle θ ₂ , the third illumination unit installation angle θ ₃ , and the fourth illumination unit installation angle θ ₄ are planes that include the inspection target surface of the abnormality detection target device. It is desirable that the angle is 30 ° to 60 ° with respect to the normal 10.
The illuminance in the device needs to be at least 500 lux. Preferably it is 700 lux or more.

[Step 1-3] This is a step of capturing a device image simultaneously with step 1-2.
A device image is taken at the same time when the abnormality detection target device is irradiated with light.
It is desirable that at least one camera that captures the abnormality detection target device and at least two directions of irradiating light. This is because there is a possibility that shadows can be formed when three-dimensional mechanical equipment is reliably photographed at different angles and is irradiated from one direction.
In addition, the angle of the camera that captures the abnormality detection target device is an angle of 0 ° to 60 ° with respect to the normal 10 of the plane including the inspection target surface of the abnormality detection target device, which suppresses the generation of shadows. In order to make the brightness of the device uniform.

When the apparatus described in FIG. 3 or FIG. 4 is used, it corresponds to the first imaging unit 21, the second imaging unit 22, and the third imaging unit 23, and includes a plane including the inspection target surface of the abnormality detection target device 1. The first imaging unit installation angle α ₁ , the second imaging unit installation angle α ₂ , and the third imaging unit installation angle α ₃ are installed with respect to the normal line 10.
Further, the first imaging unit installation angle α ₁ is not shown in FIG. 3 or FIG. 4, but α ₁ = 0 °. Thus, the first imaging unit installation angle is often parallel to the normal line 10 of the plane including the inspection target surface of the abnormality detection target device 1. The first imaging unit installation angle α ₁ , the second imaging unit installation angle α ₂ , and the third imaging unit installation angle α ₃ are preferably 0 ° to 60 °.
A normal camera can be used as the imaging unit.

[Step group 2] This is a step group constituting each pixel information of the representative image.
[Step 2-1] This is a step of searching for an abnormal image from the image group.
A device image having an abnormal part is searched from the plurality of image groups.

[Step 2-2] This is a step of configuring a post-deletion image group.
A device image having an abnormal part is deleted from the image group to form a post-deletion image group.

[Step 2-3] This is a step of constituting each pixel information of the representative image.
Based on the deleted image group, each pixel information of the representative image having the average value x _i and the standard deviation σ _i of the luminance of the corresponding pixel i of each image is configured.
A case where Step 2-3 is executed using the device image shown in FIG. 5 will be described.
FIG. 5A is a diagram depicting an image of the post-deletion image group, and FIG. 5B is a diagram depicting an image of each pixel information of the representative image, and FIG. 5A and FIG. ) Represents a pixel.
The post-deletion image group in FIG. 5A is obtained by deleting a device image having an abnormal part from the image group, and has a plurality of device images.

Therefore, for example, the average value x _i and the standard deviation σ _i are calculated for the luminance of the i-th pixel of each device image.
The information of the i-th pixel in the pixel information of each representative image in FIG. 5B is the previously calculated average value x _i and standard deviation σ _i .
Such an operation is performed for all the pixels, and each pixel information of the representative image is configured.

[Step Group 3] A step group for collecting device images at a certain time.
[Step 3-1] This is a step of spraying fluid on the abnormality detection target device.
Spray the fluid on the device to be detected. The reason why the fluid is sprayed on the abnormality detection target device is to blow off dust and dirt attached to the device. Although the environment surrounding the steel manufacturing machinery and equipment is remarkably changed and the equipment is often soiled with oil or the like, it is possible to remove the dirt by spraying a fluid and prevent the dirt from appearing as a unique part in the equipment image. The pressure of the fluid to be sprayed is a pressure higher than the atmospheric pressure, and is a pressure that does not damage the device.
The fluid is preferably sprayed immediately before the device image collection, and water, air, or the like can be used as the fluid.

[Step 3-2] This is a step of irradiating the abnormality detection target device with light from a plurality of methods.
The reason for irradiating light is to make the brightness of the device uniform. This is because the environment surrounding the steel manufacturing machine facility changes remarkably, but by irradiating light, it is possible to prevent the change in brightness around the device from appearing as a specific part of the device image.
It is desirable that the angle of irradiating light is an angle of 30 ° to 60 ° with respect to the normal line of the plane including the inspection target surface of the abnormality detection target device. This is to suppress the occurrence of shadows and make the brightness of the device uniform.
The illuminance in the device needs to be at least 500 lux. Preferably it is 700 lux or more.

[Step 3-3] This is a step of capturing a device image simultaneously with step 3-2.
A device image is taken at the same time when the abnormality detection target device is irradiated with light.
It is desirable that at least one camera that captures the abnormality detection target device and at least two directions of irradiating light. This is because there is a possibility that shadows can be formed when three-dimensional mechanical equipment is reliably photographed at different angles and is irradiated from one direction.
In addition, the angle of the camera that captures the abnormality detection target device is an angle of 0 ° to 60 ° with respect to the normal line of the plane including the inspection target surface of the abnormality detection target device, which suppresses the generation of shadows. In order to make the brightness of the device uniform.

When the apparatus described in FIG. 3 or FIG. 4 is used, it corresponds to the imaging direction 20, the first imaging unit 21, the second imaging unit 22, and the third imaging unit 23, and is the inspection target of the abnormality detection target device 1. The first imaging unit installation angle α ₁ , the second imaging unit installation angle α ₂ , and the third imaging unit installation angle α ₃ are installed with respect to the normal 10 of the plane including the surface, respectively.
Further, the first imaging unit installation angle α ₁ is not shown in FIG. 3 or FIG. 4, but α ₁ = 0 °. Thus, the first imaging unit installation angle is often parallel to the normal line 10 of the plane including the inspection target surface of the abnormality detection target device 1. The first imaging unit installation angle α ₁ , the second imaging unit installation angle α ₂ , and the third imaging unit installation angle α ₃ are preferably 0 ° to 60 °.
A normal camera can be used as the imaging unit.

[Step group 4] The device image at a certain time is determined to be abnormal.
[Step 4-1] This is a step of determining a pixel that does not satisfy a predetermined condition as an abnormal pixel.
When the luminance y _i of the pixel i of the device image is k as a real number,
x _i −kσ _i ≦ y _i ≦ x _i + kσ _i (Equation 1)
Is not satisfied, it is determined as an abnormal pixel.
When the standard deviation σ _i of the luminance of the pixel deviates from the average value x _i , for example, from −3σ _i to + 3σ _i , the pixel is determined as an abnormal pixel.

[Step 4-2] This is a step of counting the number n of abnormal pixels adjacent to each other.
The reason why the number n of abnormal pixels adjacent to each other is counted is that abnormal pixels that are not adjacent to each other are noise, and so if an abnormal image is determined based on this, so-called overdetection increases.

[Step 4-3] If n ≧ N, the image is determined as an abnormal image.
N is a predetermined numerical value, and usually 20 to 30% or more of the total number of pixels of the image is adopted.
For example, in the case of a 1000-pixel device image, if 200 to 300 pixels are adjacent and abnormal pixels, the image is determined to be an abnormal image.

[Step group 5] A step group for comprehensive determination.
As a case where the comprehensive determination unit is performed, for example, a case where an apparatus having a plurality of determination units is used is exemplified.
In the case of having a plurality of determination units, 1) a case where the image collection unit 7 and the representative image pixel information configuration unit 5 have the same determination unit 6 having different determination criteria, or 2) image collection In the case where there are a plurality of sets of the unit 7, each pixel information configuration unit 5 and the determination unit 6 of the representative image, and the like.
In FIG. 2, the comprehensive determination unit 8 is a device that performs comprehensive determination. The comprehensive determination unit 8 in FIG. 2 includes a counting unit 81, a calculation unit 82, and a final determination unit 83.
As described below, having a group of steps for performing comprehensive determination has a remarkable effect that it is possible to perform abnormality determination with less overdetection.

[Step 5-1] This is a step of counting the abnormality determination number m and the total determination number M.
The total determination number M obtained from each determination unit and the abnormality determination number m are counted.
Basically, the total number of determinations M is the number of determination units 6. However, when some of the determination units 6 are not used, the total number of determination units 6 is obtained by subtracting the number of determination units 6 that are not used from the total number of determination units 6. .

[Step 5-2] This is a step of calculating P = m / M.
P = m / M is calculated from the total determination number M and the abnormality determination number m.

[Step 5-3] This is a final determination step.
If P is greater than or equal to a predetermined value, it is finally determined as an “abnormal image”.
A value of 0.3 to 0.8 is used as the value given in advance. Desirably, a value of 0.5 to 0.7 is used.

<Configuration example>
For example, a case where “0.6” is used as a predetermined value will be described.
[When there are two judgment units]
Since there are two determination units, when M = 2 and m = 2, P is 0.6 or more. That is, only when two are determined to be abnormal images, the final determination is “abnormal image”, and when only one of them is determined to be “abnormal image”, the final determination is not “abnormal image”.
[When there are three judgment units]
Since there are two determination units, when M = 3 and m = 2, P is 0.6 or more.
That is, when any two are determined to be abnormal images, the final determination is “abnormal image”.

In the present invention, devices, sprockets, bearings, transmission shafts, frames, chains, and the like that can be used even if water or air is applied as an abnormality detection target device can be considered, but are not limited thereto. Sprockets can determine wear, chipping, deformation, bearings can determine wear, heating, transmission shaft, deformation, frame, deformation, wrinkle, chain, loosening, but not limited to these. It is not something. About the heating degree of a bearing, it can determine effectively by acquiring a color image by imaging. It is more effective if water or air is applied to the abnormality detection target device, and then air (gas) is blown to blow off moisture.
[Example 1]
A case where there is one imaging unit will be described with reference to FIG.
FIG. 6 (a) is an example in which the present invention is applied to one sprocket located on the side of the exhausting section of the sintering machine. In FIG. 6A, 1 is a sprocket that is an abnormality detection target device, 21 is a first imaging unit, 31 is a first fluid spraying unit, and 41 is a first illumination unit. The sprocket 1 rotates around the rotation axis O, and when the sprocket 1 reaches a predetermined angle, water is sprayed from the first fluid spraying part 31 and the first illumination part 41 emits light. When irradiated, the first imaging unit 21 captures an image of the sprocket 1. Using the average value x _i and the standard deviation σ _i of the luminance of each pixel obtained from the image of the sprocket photographed at normal time, the photographed image is determined as k = 3 in (Equation 1),
x _i −3σ _i ≦ y _i ≦ x _i + 3σ _i (Formula 3)
I went from.

[Example 2]
A case where there are two imaging units will be described with reference to FIG.
FIG. 6B is also an example in which the present invention is applied to one of the sprockets located on the side of the discharge portion of the sintering machine. In FIG. 6B, 1 is a sprocket that is an abnormality detection target device, 21 is a first imaging unit, 22 is a second imaging unit, 31 is a first fluid spraying unit, 32 is a second fluid spraying unit, and 41. Is a first illumination unit, and 42 is a second illumination unit. The sprocket 1 rotates around the rotation axis O, and when the sprocket 1 reaches a predetermined angle, steam is sprayed from the first fluid spraying portion and the second fluid spraying portion, When the first illumination unit 41 and the second illumination unit 42 emit light, the first imaging unit 21 and the second imaging unit 22 capture an image of the sprocket 1. Using the average value x _i and the standard deviation σ _i of the luminance of each pixel obtained from the image of the sprocket photographed at normal time, the photographed image is determined as k = 3 in (Equation 1),
x _i −3σ _i ≦ y _i ≦ x _i + 3σ _i (Formula 3)
I went from.

[Test results]
In the evaluation of the results of experiments on the present invention example using the method of the present invention, the detection rate was 90% or more and the overdetection rate was 5% or less.
In Example 1 of the present invention, the detection rate was 95% with respect to 100 abnormal samples, and the overdetection rate was 4% with respect to 100 normal samples.
In Invention Example 2, the detection rate was 98% with respect to 100 abnormal samples, and the overdetection rate was 3% with respect to 100 normal samples.
Thus, in the method of the present invention, the detection rate can be 90% or more, the overdetection rate can be 5% or less, and can be put to practical use.

  It can be used to determine the abnormality of machinery and equipment in a poor environment.

DESCRIPTION OF SYMBOLS 1 ... Abnormality detection object apparatus, 10 ... Normal of plane containing inspection object surface of abnormality detection object apparatus, 2 ... Imaging part, 20 ... Imaging direction, 21 ... 1st imaging part, 22 ... 2nd imaging part, 23 ... Third imaging unit, α ₁ ... first imaging unit installation angle, α ₂ ... second imaging unit installation angle, α ₃ ... third imaging unit installation angle, 3 ... fluid spraying unit, 30 ... fluid spraying direction, 31 ... 1st fluid spraying part, 32 ... 2nd fluid spraying part, 4 ... Illumination part, 40 ... Light irradiation direction, 41 ... 1st illumination part, 42 ... 2nd illumination part, 43 ... 3rd illumination part, 44 ... 4th illumination part, (theta) ₁ ... 1st illumination part installation angle, (theta) ₂ ... 2nd illumination part installation angle, (theta) ₃ ... 3rd illumination part installation angle, (theta) ₄ ... 4th illumination part installation angle, 5 ... Representative Each pixel information constituent unit, 6 ... determination unit, 7 ... image collection unit, 8 ... comprehensive determination unit, 81 ... counting unit, 82 ... calculation unit, 83 ... final determination unit, 101 ... sintering machine sp Rocket, 102B ... One sprocket, 104 ... Sintered bullet trolley (sintered bullet), 104C ... Support stand, 109 ... Crusher, 110 ... Crushing guide, 120 ... Main body wall, 130 ... Guide part, O ... Rotating shaft, H ... Horizontal line, S ... Sinter cake that has fallen, γ ... Angle, G ... Minimum gap between the sintered bullet trolley 104 and the main body wall 120.

Claims (18)

A method for detecting an abnormality of equipment provided in a steel manufacturing machine facility,
An image of an abnormality detection target device (hereinafter referred to as “device image”)
Spraying fluid on the anomaly detection target device;
Irradiating the abnormality detection target device with light from a plurality of different directions;
Capturing an image of the abnormality detection target device simultaneously with irradiating the abnormality detection target device with light from a plurality of different directions; and
A group of steps taken through
From a plurality of the device images (hereinafter referred to as “image group”),
Searching for a device image (hereinafter referred to as “abnormal image”) having an abnormal image portion (hereinafter referred to as “abnormal portion”);
Deleting abnormal images from the image group and configuring the deleted image group;
A group of steps constituting the pixel information of the representative image having the average value x _i and the standard deviation σ _i of the corresponding pixel i in each image of the post-deletion image group;
A device image of the abnormality detection target device at a certain time (hereinafter referred to as a “device image at a certain time”),
Spraying fluid on the anomaly detection target device;
Irradiating light from a plurality of different directions to the abnormality detection target device and simultaneously collecting the device image;
A group of steps obtained through
The abnormality determination of the device image at the certain time
When the luminance y _i of the pixel i of the device image is k as a real number,
x _i −kσ _i ≦ y _i ≦ x _i + kσ _i (in the case where (Equation 1) is not satisfied)
Counting the number n of pixels determined to be abnormal pixels and adjacent to each other;
Determining that the image is an abnormal image when n is a predetermined number N or more;
Carried out through,
The abnormality detection method for mechanical equipment, wherein the abnormality detection target device is a sintering machine .   The method for detecting an abnormality in a mechanical facility according to claim 1, wherein the real number k is a real number of 1 or more. When performing the determination step multiple times,
Counting the total number of determinations in the determination step (hereinafter referred to as “total number of determinations”) M and the total number of determinations determined in the determination step as abnormal images (hereinafter referred to as “abnormality determination numbers”) m; ,
From the total determination number M and the abnormality determination number m for the abnormality detection target device, P = m / M (Expression 2)
A step of calculating
The abnormality detection of the machine equipment according to claim 1 or 2, further comprising: a step group including a step of finally determining an abnormal image when P is equal to or greater than a predetermined value. Method. In the case of having a plurality of imaging units that capture images of the abnormality detection target device simultaneously with the step of irradiating light from a plurality of different directions to the abnormality detection target device,
A step group constituting an image group for each imaging unit;
In each image group, a step group that configures a post-deletion image group in which an abnormal image is deleted and configures pixel information of each representative image corresponding to each post-deletion image group;
The method according to claim 3, further comprising:   The angle between the direction in which the image of the abnormality detection target device is captured and the normal line of the plane including the inspection target surface of the abnormality detection target device is 0 ° to 60 °, and the direction of irradiating the light and the abnormality detection target device The mechanical equipment abnormality detection method according to any one of claims 1 to 4, wherein an angle formed with a normal line of a plane including the surface to be inspected is 30 ° to 60 °.   The method for detecting an abnormality of a mechanical facility according to any one of claims 1 to 5, wherein the fluid is water.   The method for detecting an abnormality in a mechanical facility according to any one of claims 1 to 5, wherein the fluid is air and water.   The method for detecting an abnormality of a mechanical facility according to claim 6 or 7, wherein a gas is further sprayed after the fluid is sprayed on the abnormality detection target device.   An optical filter that uses an imaging device having detection sensitivity in a wavelength band of 1.1 μm or less as the imaging device and cuts light having a wavelength of 0.7 μm or less in the imaging device in the step of capturing an image of the abnormality detection target device The method for detecting an abnormality of a mechanical facility according to any one of claims 6 to 8, wherein an optical filter that cuts light having a wavelength of 0.95 to 1.0 µm is attached in an overlapping manner. An apparatus for detecting an abnormality of equipment provided in a steel manufacturing machine facility,
For each abnormality detection target device,
A fluid spraying part for spraying fluid on the abnormality detection target device;
An illumination unit that emits light from a plurality of different directions to the abnormality detection target device;
An imaging unit that collects an image of the abnormality detection target device (hereinafter, referred to as “device image”) at the same time that the illumination unit irradiates light;
A representative image pixel information constituting unit constituting representative image pixel information for a plurality of device images (hereinafter referred to as “image group”);
A determination unit that determines whether there is an abnormality in a device image at a certain time of the abnormality detection target device (hereinafter referred to as a “device image at a certain time”);
Have
The representative image each pixel information constituting unit,
A post-deletion image group constituting unit that deletes a device image (hereinafter, referred to as “abnormal image”) having an abnormal image part (hereinafter referred to as “abnormal part”) from the image group and constitutes a post-deletion image group; ,
An average value x _i of the luminance of the corresponding pixel i of each image in the post-deletion image group and a standard image luminance component that is used as each pixel information of the standard deviation σ _i representative image;
Have
The determination unit
When the luminance y _i of the pixel i of the device image at a certain time is k as a real number,
x _i −kσ _i ≦ y _i ≦ x _i + kσ _i ... (Expression 1)
Means for counting the number n of pixels determined to be abnormal pixels and adjacent to each other;
Means for determining that the image is an abnormal image when n is a predetermined number N or more;
I have a,
The abnormality detection apparatus for machine equipment, wherein the abnormality detection target device is a sintering machine . The apparatus according to claim 10 , wherein the real number k is a real number equal to or greater than one. A plurality of the determination units;
A counting unit that counts the total number of determinations (hereinafter referred to as “total number of determinations”) M and the total number of determinations determined as abnormal images (hereinafter referred to as “abnormality determination numbers”) m in the determination unit. When,
From the total determination number M and the abnormality determination number m, P = m / M (Expression 2)
A calculation unit for calculating
If the P is equal to or greater than a predetermined value, a final determination unit that finally determines an abnormal image;
Abnormality detecting device of a machine installation according to claim 10 or claim 11, characterized in that it further comprises a comprehensive determination unit having a. In the case of having a plurality of imaging units that irradiate light from a plurality of different directions to the abnormality detection target device and simultaneously capture an image of the abnormality detection target device,
An image collection unit that constitutes an image group for each imaging unit;
Each image group further includes a representative image pixel information configuration unit that configures a post-deletion image group in which an abnormal image is deleted and configures the representative image pixel information corresponding to each post-deletion image group. The abnormality detection apparatus for mechanical equipment according to claim 12 . The angle between the direction in which the image of the abnormality detection target device is captured and the normal line of the plane including the inspection target surface of the abnormality detection target device is 0 ° to 60 °, and the direction of irradiating the light and the abnormality detection target device The mechanical equipment abnormality detection device according to any one of claims 10 to 13, wherein an angle formed with a normal line of a plane including the surface to be inspected is 30 ° to 60 °. 15. The abnormality detection device for mechanical equipment according to any one of claims 10 to 14 , wherein the fluid is water. The mechanical fluid abnormality detection device according to any one of claims 10 to 14 , wherein the fluid is air and water. The mechanical equipment abnormality detection device according to claim 15 or 16 , wherein after the fluid is sprayed on the abnormality detection target device, gas is further sprayed. An optical filter that uses an imaging device having detection sensitivity in a wavelength band of 1.1 μm or less as the imaging device and cuts light having a wavelength of 0.7 μm or less in the imaging device in the step of capturing an image of the abnormality detection target device The mechanical equipment abnormality detection device according to any one of claims 15 to 17 , wherein an optical filter that cuts light having a wavelength of 0.95 to 1.0 µm is attached in an overlapping manner.

Images