JPH11129083A - Cutting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately grasp and control a cutting condition of a cutting portion generating lights and high heat by installing a macrophotographing means to take the image near the cutting portion, or a macrophotographing and a microphotographing means, and a image processing part processing the taken image information in a gas, laser, or plasma cutting device. SOLUTION: A macrocamera 11 at the tip part of an arm 11a mounted to a movable laser torch 7 being connected to a laser oscillator 6 by a bellows optical path 8, photographs the proximity of the cutting portion, and the image information is made into multiple thresholdlevels at an image processing part 22. Furthermore, the image information of a molten pool inside the cutting portion taken by a microcamera being arranged beside a lens inside a laser torch 7, is made into multiple thresholdlevels and pattern matching treated at the image processing part 22. A control part 21 controls the cutting speed, laser light strength, pulse and the like of a cutting device 1 based on the judgement of the image processing part 22 so as to always carry out the optimum cutting automatically.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cutting apparatus using gas, plasma, or laser, and more particularly to a cutting apparatus which monitors a cutting portion to judge and control a cutting state.

[0002]

2. Description of the Related Art When cutting with a gas, plasma or laser cutting device, cutting is performed by burning and melting a material to be cut by heat or air current, and removing generated slag. Hereinafter, a laser cutting apparatus will be described as an example with reference to the drawings. FIG. 8 is a cross-sectional view showing the vicinity of a cutting portion of a conventional laser cutting device, and FIG. 9 is a perspective view showing the vicinity of a cutting portion.

FIG. 8A shows a state in which cutting is normally performed, and a material to be cut 51 is cut by a laser torch 50. At this time, the molten pool 51a in which the material to be cut 51 is melted is generated inside the material to be cut 51, and is burned and removed, whereby cutting proceeds.

However, as shown in FIG. 8 (b), when a cutting defect occurs and the laser does not penetrate the material to be cut 51, the molten slag 51b blows out to the surface of the material to be cut, and a so-called burning phenomenon occurs. I do. When such a situation occurs, not only the cutting is incomplete, but also the quality of the surface and cut surface of the cut material 51 deteriorates.

Further, as shown in FIG. 9, a cut material 52 cut from the cut material 51 is caught without falling, one end of the cut material 52 is inclined below the cut material 51, and the other end projects above the cut material 51. May be. When the laser torch 50 attempts to pass through a certain position of the cutting material 52 in such a state, the tip of the laser torch 50 may come into contact with the cutting material 52 and be damaged.

In order to avoid such cutting failures and failures, conventionally, an operator has always monitored a cut portion.
When the burning phenomenon occurs, the cutting is immediately stopped and the cutting is restarted.
When there was, the laser torch 50 was stopped and bypassed.

[0007]

However, since the cutting site emits strong light and high heat, it is difficult to accurately grasp the cutting state, and it is difficult to work for a long time due to fatigue of the worker. Because of the difficulty, there was a problem that many replacement personnel were required.

[0008]

In order to solve the above-mentioned problems, a cutting apparatus according to the present invention includes a gas, laser, or plasma cutting apparatus, and a macro photographing means for acquiring an image near a cut portion; An image processing unit for converting the image information obtained by the macro image capturing means into a multi-threshold, wherein the image processing unit is configured to be able to control the cutting device based on a processing result of the image processing unit. Preferably, the image information is subjected to a pattern matching process on the multi-threshold image information.

The cutting device is a laser cutting device, and a laser torch is provided with micro-photographing means for photographing a cut portion from above, and image information obtained by the micro-photographing means is processed by the image processing unit at a multi-threshold. And a pattern matching process, and the cutting device can be controlled based on a processing result of the image processing unit.

The micro photographing means may be built in the laser torch or have a photographing hole on a side surface of the laser torch, and the micro photographing means may be arranged outside the laser torch and opposed to the photographing hole. And a half split is arranged inside the laser torch at an angle of 45 ° with the photographing hole and the laser light.

Preferably, the pattern matching process is a neural network.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION

[First Embodiment] An embodiment of a cutting device according to the present invention will be described with reference to the drawings, taking a laser cutting device as an example.
FIG. 1 is an overall perspective view of a laser cutting apparatus according to the present embodiment, FIG. 2 is a cross-sectional view of a laser torch, FIG. 3 is a cross-sectional view showing the vicinity of a cut portion of the laser cutting apparatus, and FIG. FIG. 5 is a diagram illustrating image processing of the micro camera, FIG. 6 is a diagram illustrating a neural network, and FIG. 7 is a diagram illustrating another configuration of the laser torch.

The laser cutting device 1 according to the present embodiment
Are movably installed on two rails 2 laid in parallel. On the frame 3 of the cutting device 1, a laser oscillator 6, a control panel 4 for setting the cutting device 1, and an operation panel 5 for instructing an operation are provided, and according to an instruction from the operation panel 5 or automatically. Control unit for controlling the cutting device 1
21, an image processing unit 22 described later and the like are mounted.

A laser torch 7 is mounted on the front surface of the frame 3 and is movably mounted in the direction perpendicular to the rail 2 along the frame 3. By adjusting the movement of the laser torch 7 and the movement of the frame 3, an arbitrary position of the workpiece 10 on the gantry 9 laid between the two rails 2 can be cut.

The laser torch 7 and the laser oscillator 6 are connected by an optical path 8 constituted by a bellows. The optical path 8 is provided with a constant optical path length device 8a for reversing the traveling direction of the laser light, and is configured to move half the distance with the movement of the laser torch 7. As a result, the entire length of the optical path 8 is always kept constant, and high-quality cutting can be performed while keeping the energy and the focus of the laser light constant.

An arm 11a is attached to the laser torch 7, and a macro camera 11 is attached to a tip of the arm 11a. The macro camera 11 is located above the tip of the laser torch 7 so as not to be affected by heat due to cutting,
The cut site is arranged to be photographed from obliquely above.

As shown in FIG. 2, a lens 12 is attached to the inner end of the laser torch 7. This lens
Laser light supplied from laser oscillator 6 by 12
13 can be converged, and the material to be cut 10 can be cut by combustion.

The lens inside the laser torch 7 and
A micro camera 14 is provided on the side of 12, and is configured to photograph a cut portion from a photographing hole 15 provided at the tip of the laser torch 7. With this configuration, the micro camera 14 can photograph the cut portion from almost immediately above, and the molten pool 10a formed inside the workpiece 10 is cut.
Can be taken.

When cutting and moving the laser torch 7, the macro camera 11 and the micro camera 14 always take pictures and perform image processing described later. Here, as shown in FIG. 3, when a cutting failure occurs in which the laser beam does not penetrate the workpiece 10, a so-called burning phenomenon occurs in which the slag 10b is ejected onto the surface of the workpiece 10. FIG. 3 also shows a cutting material 16 that does not fall due to being caught.

At this time, the image photographed by the macro camera 11 is as shown in FIG. That is, the portion of the material to be cut 10 is dark, and the portion of the slag 10b due to the burning phenomenon is extremely bright. Further, since the angle of the cut material 16 is different, the portions have different brightness. Therefore, this image is divided into m, and each position, that is, the scan layers S1 to Sm
Scan at to obtain a luminance level.

FIG. 4B is an example of a three-dimensional graph showing the brightness level V of the photographed image. The brightness level of the cut material 10 is V1, the brightness level of the cut material 16 is V2, and the brightness level of the slag 10b. V3. In order to detect the presence of the slug 10b or the cut material 16 in the macro camera 11, the luminance level V is set to multiple thresholds, a threshold P1 is set between the luminance levels V1 and V2, and the luminance levels V2 and A threshold value P2 is set between the level and the level V3.

When a brightness level equal to or higher than the threshold value P2 is detected from the obtained photographed image, it is determined that a burning phenomenon has occurred and the cutting is immediately stopped. To resume. When a brightness level equal to or more than the threshold value P1 and equal to or less than the threshold value P2 is detected, it is determined that the cutting material 16 is on the path of the laser torch 7, and the progress of the laser torch 7 can be stopped or bypassed.

The image photographed by the micro camera 14 is
As shown in FIG. 5A, the molten pool 10a inside the workpiece 10 can be captured. This image is scanned on the scan layers S1 to Sm in the same manner as the image obtained by the macro camera 11, and a luminance level V3 is obtained. Then, as shown in FIG. 5 (b), a threshold value P3 is set, and a brightness level V3 larger than the threshold value P3 is detected, thereby obtaining the coordinates X of the shape of the molten pool 10a.
Can be recognized. The recognized coordinates X are sent to the image processing unit 22 and evaluated by the pattern matching method. In the present embodiment, a neural network (hereinafter simply referred to as NNW) is used as the pattern matching method.

Since the rear side of the molten pool 10a is concavely curved, the coordinates X of the shape obtained for one scan layer S are obtained.
Is a maximum of four. Therefore, the coordinates X obtained from the scan layer S1 are X11 to X14, the coordinates X obtained from the scan layer S2 are X21 to X24, and the coordinates X obtained from the scan layer Sm are Xm1 to Xm4. The data of all the coordinates X obtained from one image is collectively called a learning pattern,
The shape of the molten pool 10a is recognized based on this learning pattern.

The NNW according to the present embodiment has three neuron layers as shown in FIG. 6, and comprises an input layer A, an intermediate layer B, and an output layer C. The back propagation method is used for the learning rule of the NNW, the batch correction method is used for the correction method of the load W, and the moment method is used for calculation of the load correction amount. The batch correction method is a method in which the total amount of correction is calculated for the entire learning pattern, and the load is corrected using the total.The number of corrections is smaller than the sequential correction method in which correction is performed for each learning pattern individually, so that learning is performed. Time can be reduced.

A learning pattern is input to the input layer A and the output layer C, and the NNW is learned. The intermediate layer B is given the sum of the data of the input layer A with the load W. Each neuron in the input layer A connects to two neurons in the intermediate layer B (which mimics the synaptic connections of the organism). That is, the intermediate layer has elements B1 to B2m corresponding to the scan layers S1 to Sm. In this way, by connecting the neurons of the input layer A to the two neurons of the intermediate layer B, a nonlinear system can be constructed.

For example, the index of each coordinate X in a certain scan layer Sm is i, and the index of the intermediate layer B is j (j
= 2m or 2m-1), the load imposed between the coordinate Xmi and the intermediate layer Bj is expressed as Wij. Using this, the intermediate layer Bj is represented by the following equation.

[0028]

(Equation 1)

Here, the load Wij is corrected by the batch correction method as described above, and the load correction amount ΔW (k) at that time is calculated by the moment method. Specifically, assuming that the previous load correction amount is ΔW (k−1) and the correction amount from the error is d, the load correction amount is obtained by the following equation.

[0030]

(Equation 2)

Here, a is a stabilization coefficient.

Each neuron of the intermediate layer B converts the given sum into a real number or a range of 0 to 1 by a sigmoid function and outputs the result. The output of the intermediate layer B is, as described above,
This is given to the output layer C as a weighted sum.

Here, in the back propagation learning method, the correction amount ΔVkj of the synaptic load Vkj connecting the intermediate layer and the output layer and the correction amount Δγk of the threshold value γk of the neuron of the output layer are given by the following equations.

[0034]

(Equation 3)

The teacher signal Tk is 1.0 when the learning pattern is a good pattern, and 0.0 when the learning pattern is good. Thus, the difference Δ between the output Ok of the output layer and the teacher signal Tk
By decreasing V, the NNW learns and a more correct output layer output Ok can be obtained. Finally, the output layer C converts the sum given from the intermediate layer B into a range of 0 to 1 by a sigmoid function and outputs the converted result. If not, it is not.

Based on the judgment of the image processing unit 22 by the NNW, the control unit 21 controls the cutting device 1 to control the cutting speed, the intensity of the laser beam, the pulse of the laser beam, and the like. In this way, by always monitoring the cutting state and responding to the situation, it is possible to always perform the optimum cutting.

Although the micro camera 14 is provided inside the laser torch 7 in the above embodiment, it may be difficult to incorporate the laser torch 7 depending on the size and structure of the laser torch 7. FIG. 7 shows an example in which the micro camera 14 is attached to the outside of the laser torch 7, and the same reference numerals are given to the same portions as those in the above embodiment, and the description is omitted.

The laser torch 17 shown in FIG. 7 has a photographing hole 20 on the side surface, and the micro camera 14 is arranged at a position facing the photographing hole 20 so as to photograph the inside of the laser torch 17. Further, a half split 18 is arranged inside the laser torch 17 so as to have an angle of 45 ° with respect to the laser beam 13 and the imaging hole 20.

As a result, the laser beam 13 supplied from the laser oscillator 6 is split into the half split 18 and the lens 12.
To burn the material to be cut 10. Then, the light emitted from the molten pool 10a at the cut portion passes through the lens 12 and becomes parallel to the laser light 13, is reflected by the half split 18, and is photographed by the micro camera 14. At this time, since the laser light 13 is an invisible light,
14 will not be filmed.

[Other Embodiments] In each of the above embodiments, a laser cutting apparatus has been described as an example. However, the present invention is not limited to this, and is applicable to a gas or plasma cutting apparatus. is there. However, in the gas or plasma cutting apparatus, the molten pool can be observed from the outside, so that there is no need to provide a micro camera, and the cutting state can be monitored only by the macro camera.

Since a gas flame or a plasma arc emits visible light, it is also possible to directly recognize not only the shape of the molten pool but also the shape of the gas flame or the like and determine the properties thereof by a pattern matching method. In addition, it is possible to confirm ignition.

It is to be noted that, for example, a slag 10b of a burning phenomenon, a cut material 16, and a gas flame 3 are generated by one macro camera.
When recognizing one, three thresholds may be set and these may be recognized.

In each of the above embodiments, an image is taken by an optical camera. However, it is sufficient that an image to be processed can be obtained. For example, an electronic scanning image, an ultrasonic scanning image, and a thermograph An image, an X-ray image, or the like may be used.

[0044]

In the cutting apparatus according to the present invention, the cutting conditions in the gas cutting, the plasma arc cutting, and the laser cutting can be monitored by the above-described configuration. , The optimum cutting can always be performed. Further, when a cutting failure such as a burning phenomenon occurs, the cutting can be immediately stopped, and there is no possibility that the quality of the cut portion is deteriorated.

Further, since there is no need for monitoring, it is possible to relieve the worker from labor and to reduce the number of personnel who spend the work for the user. In addition, the cutting quality can be kept constant without the judgment of the cutting state depending on the subjectivity of each worker, and the cutting efficiency can be increased by automatically processing a fault that has occurred during the cutting.

[Brief description of the drawings]

FIG. 1 is an overall perspective view of a laser cutting device according to a first embodiment.

FIG. 2 is a schematic sectional view of a laser torch.

FIG. 3 is a cross-sectional view showing the vicinity of a cut portion of the laser cutting device.

FIG. 4 is a diagram illustrating image processing of a macro camera.

FIG. 5 is a diagram illustrating image processing of a micro camera.

FIG. 6 is a diagram illustrating a neural network.

FIG. 7 is a diagram showing another configuration of the laser torch.

FIG. 8 is a cross-sectional view showing the vicinity of a cutting portion of a conventional laser cutting device.

FIG. 9 is a perspective view showing the vicinity of a cut portion.

[Explanation of symbols]

 P: threshold value S: scan layer V: brightness level W: load 1: laser cutting device 2: rail 3: frame 4: control box 5: operation panel 6: laser oscillator 7: laser torch 8: optical path 8a: constant optical path length device 9 ... stand 10 ... material to be cut 10a ... molten pool 10b ... slug 11 ... macro camera 11a ... arm 12 ... lens 13 ... laser light 14 ... micro camera 15 ... shooting hole 16 ... cutting material 17 ... laser torch 18 ... half split 19 … Reflected light 20… Shooting hole 21… Control unit 22… Image processing unit

Continued on the front page (51) Int.Cl. ⁶ Identification symbol FI G06T 7/00 G06F 15/62 400 (72) Inventor Kim Bing-cheol 3-35-16 Nishi Koiwa, Edogawa-ku, Tokyo Koike Oxygen Industry Co., Ltd. Inside

Claims (6)

[Claims] In a gas, laser or plasma cutting apparatus, there is provided a macro photographing means for acquiring an image in the vicinity of a cut part, and an image processing unit for multiplying the image information acquired by the macro photographing means into a multi-threshold. A cutting device configured to control the cutting device based on a processing result of the image processing unit. 2. The cutting apparatus according to claim 1, wherein the image processing unit further performs a pattern matching process on the image information that has been subjected to the multiple thresholding. 3. The laser cutting device according to claim 1, wherein the laser torch is provided with a micro-photographing means for photographing a cut portion from above, and image information obtained by the micro-photographing means is multiplexed by the image processing unit. 2. The cutting device according to claim 1, wherein the cutting device is configured to perform thresholding and pattern matching processing and control the cutting device based on a processing result of the image processing unit. 4. The cutting apparatus according to claim 3, wherein said micro photographing means is an optical camera built in said laser torch. 5. The laser torch has a photographing hole on a side surface, and the micro photographing means is arranged outside the laser torch and opposed to the photographing hole, and the photographing hole and the laser light are arranged inside the laser torch. The cutting device according to claim 3, wherein the half split is disposed at an angle of 45 °. 6. The cutting apparatus according to claim 1, wherein the pattern matching processing is a neural network.